net_sched: SFB flow scheduler
[linux-2.6-block.git] / net / socket.c
CommitLineData
1da177e4
LT
1/*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
02c30a84 7 * Ross Biro
1da177e4
LT
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
89bddce5 45 * Tigran Aivazian : Made listen(2) backlog sanity checks
1da177e4
LT
46 * protocol-independent
47 *
48 *
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
53 *
54 *
55 * This module is effectively the top level interface to the BSD socket
89bddce5 56 * paradigm.
1da177e4
LT
57 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
1da177e4 61#include <linux/mm.h>
1da177e4
LT
62#include <linux/socket.h>
63#include <linux/file.h>
64#include <linux/net.h>
65#include <linux/interrupt.h>
aaca0bdc 66#include <linux/thread_info.h>
55737fda 67#include <linux/rcupdate.h>
1da177e4
LT
68#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
4a3e2f71 71#include <linux/mutex.h>
1da177e4
LT
72#include <linux/wanrouter.h>
73#include <linux/if_bridge.h>
20380731
ACM
74#include <linux/if_frad.h>
75#include <linux/if_vlan.h>
1da177e4
LT
76#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
1da177e4
LT
81#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
3ec3b2fb 86#include <linux/audit.h>
d86b5e0e 87#include <linux/wireless.h>
1b8d7ae4 88#include <linux/nsproxy.h>
1fd7317d 89#include <linux/magic.h>
5a0e3ad6 90#include <linux/slab.h>
1da177e4
LT
91
92#include <asm/uaccess.h>
93#include <asm/unistd.h>
94
95#include <net/compat.h>
87de87d5 96#include <net/wext.h>
f8451725 97#include <net/cls_cgroup.h>
1da177e4
LT
98
99#include <net/sock.h>
100#include <linux/netfilter.h>
101
6b96018b
AB
102#include <linux/if_tun.h>
103#include <linux/ipv6_route.h>
104#include <linux/route.h>
6b96018b
AB
105#include <linux/sockios.h>
106#include <linux/atalk.h>
107
1da177e4 108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
027445c3
BP
109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
112 unsigned long nr_segs, loff_t pos);
89bddce5 113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
1da177e4
LT
114
115static int sock_close(struct inode *inode, struct file *file);
116static unsigned int sock_poll(struct file *file,
117 struct poll_table_struct *wait);
89bddce5 118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
89bbfc95
SP
119#ifdef CONFIG_COMPAT
120static long compat_sock_ioctl(struct file *file,
89bddce5 121 unsigned int cmd, unsigned long arg);
89bbfc95 122#endif
1da177e4 123static int sock_fasync(int fd, struct file *filp, int on);
1da177e4
LT
124static ssize_t sock_sendpage(struct file *file, struct page *page,
125 int offset, size_t size, loff_t *ppos, int more);
9c55e01c 126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
c6d409cf 127 struct pipe_inode_info *pipe, size_t len,
9c55e01c 128 unsigned int flags);
1da177e4 129
1da177e4
LT
130/*
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
133 */
134
da7071d7 135static const struct file_operations socket_file_ops = {
1da177e4
LT
136 .owner = THIS_MODULE,
137 .llseek = no_llseek,
138 .aio_read = sock_aio_read,
139 .aio_write = sock_aio_write,
140 .poll = sock_poll,
141 .unlocked_ioctl = sock_ioctl,
89bbfc95
SP
142#ifdef CONFIG_COMPAT
143 .compat_ioctl = compat_sock_ioctl,
144#endif
1da177e4
LT
145 .mmap = sock_mmap,
146 .open = sock_no_open, /* special open code to disallow open via /proc */
147 .release = sock_close,
148 .fasync = sock_fasync,
5274f052
JA
149 .sendpage = sock_sendpage,
150 .splice_write = generic_splice_sendpage,
9c55e01c 151 .splice_read = sock_splice_read,
1da177e4
LT
152};
153
154/*
155 * The protocol list. Each protocol is registered in here.
156 */
157
1da177e4 158static DEFINE_SPINLOCK(net_family_lock);
190683a9 159static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
1da177e4 160
1da177e4
LT
161/*
162 * Statistics counters of the socket lists
163 */
164
c6d409cf 165static DEFINE_PER_CPU(int, sockets_in_use);
1da177e4
LT
166
167/*
89bddce5
SH
168 * Support routines.
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
1da177e4
LT
171 */
172
1da177e4
LT
173/**
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
178 *
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 */
183
230b1839 184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
1da177e4 185{
230b1839 186 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
1da177e4 187 return -EINVAL;
89bddce5 188 if (ulen == 0)
1da177e4 189 return 0;
89bddce5 190 if (copy_from_user(kaddr, uaddr, ulen))
1da177e4 191 return -EFAULT;
3ec3b2fb 192 return audit_sockaddr(ulen, kaddr);
1da177e4
LT
193}
194
195/**
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
201 *
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
206 * accessible.
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
210 */
89bddce5 211
11165f14 212static int move_addr_to_user(struct sockaddr *kaddr, int klen,
213 void __user *uaddr, int __user *ulen)
1da177e4
LT
214{
215 int err;
216 int len;
217
89bddce5
SH
218 err = get_user(len, ulen);
219 if (err)
1da177e4 220 return err;
89bddce5
SH
221 if (len > klen)
222 len = klen;
230b1839 223 if (len < 0 || len > sizeof(struct sockaddr_storage))
1da177e4 224 return -EINVAL;
89bddce5 225 if (len) {
d6fe3945
SG
226 if (audit_sockaddr(klen, kaddr))
227 return -ENOMEM;
89bddce5 228 if (copy_to_user(uaddr, kaddr, len))
1da177e4
LT
229 return -EFAULT;
230 }
231 /*
89bddce5
SH
232 * "fromlen shall refer to the value before truncation.."
233 * 1003.1g
1da177e4
LT
234 */
235 return __put_user(klen, ulen);
236}
237
e18b890b 238static struct kmem_cache *sock_inode_cachep __read_mostly;
1da177e4
LT
239
240static struct inode *sock_alloc_inode(struct super_block *sb)
241{
242 struct socket_alloc *ei;
eaefd110 243 struct socket_wq *wq;
89bddce5 244
e94b1766 245 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
1da177e4
LT
246 if (!ei)
247 return NULL;
eaefd110
ED
248 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
249 if (!wq) {
43815482
ED
250 kmem_cache_free(sock_inode_cachep, ei);
251 return NULL;
252 }
eaefd110
ED
253 init_waitqueue_head(&wq->wait);
254 wq->fasync_list = NULL;
255 RCU_INIT_POINTER(ei->socket.wq, wq);
89bddce5 256
1da177e4
LT
257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
1da177e4
LT
262
263 return &ei->vfs_inode;
264}
265
43815482 266
ff0c7d15 267
43815482
ED
268static void wq_free_rcu(struct rcu_head *head)
269{
270 struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
271
272 kfree(wq);
273}
274
1da177e4
LT
275static void sock_destroy_inode(struct inode *inode)
276{
43815482 277 struct socket_alloc *ei;
eaefd110 278 struct socket_wq *wq;
43815482
ED
279
280 ei = container_of(inode, struct socket_alloc, vfs_inode);
eaefd110
ED
281 wq = rcu_dereference_protected(ei->socket.wq, 1);
282 call_rcu(&wq->rcu, wq_free_rcu);
43815482 283 kmem_cache_free(sock_inode_cachep, ei);
1da177e4
LT
284}
285
51cc5068 286static void init_once(void *foo)
1da177e4 287{
89bddce5 288 struct socket_alloc *ei = (struct socket_alloc *)foo;
1da177e4 289
a35afb83 290 inode_init_once(&ei->vfs_inode);
1da177e4 291}
89bddce5 292
1da177e4
LT
293static int init_inodecache(void)
294{
295 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
89bddce5
SH
296 sizeof(struct socket_alloc),
297 0,
298 (SLAB_HWCACHE_ALIGN |
299 SLAB_RECLAIM_ACCOUNT |
300 SLAB_MEM_SPREAD),
20c2df83 301 init_once);
1da177e4
LT
302 if (sock_inode_cachep == NULL)
303 return -ENOMEM;
304 return 0;
305}
306
b87221de 307static const struct super_operations sockfs_ops = {
c6d409cf
ED
308 .alloc_inode = sock_alloc_inode,
309 .destroy_inode = sock_destroy_inode,
310 .statfs = simple_statfs,
1da177e4
LT
311};
312
c23fbb6b
ED
313/*
314 * sockfs_dname() is called from d_path().
315 */
316static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
317{
318 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
319 dentry->d_inode->i_ino);
320}
321
3ba13d17 322static const struct dentry_operations sockfs_dentry_operations = {
c23fbb6b 323 .d_dname = sockfs_dname,
1da177e4
LT
324};
325
c74a1cbb
AV
326static struct dentry *sockfs_mount(struct file_system_type *fs_type,
327 int flags, const char *dev_name, void *data)
328{
329 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
330 &sockfs_dentry_operations, SOCKFS_MAGIC);
331}
332
333static struct vfsmount *sock_mnt __read_mostly;
334
335static struct file_system_type sock_fs_type = {
336 .name = "sockfs",
337 .mount = sockfs_mount,
338 .kill_sb = kill_anon_super,
339};
340
1da177e4
LT
341/*
342 * Obtains the first available file descriptor and sets it up for use.
343 *
39d8c1b6
DM
344 * These functions create file structures and maps them to fd space
345 * of the current process. On success it returns file descriptor
1da177e4
LT
346 * and file struct implicitly stored in sock->file.
347 * Note that another thread may close file descriptor before we return
348 * from this function. We use the fact that now we do not refer
349 * to socket after mapping. If one day we will need it, this
350 * function will increment ref. count on file by 1.
351 *
352 * In any case returned fd MAY BE not valid!
353 * This race condition is unavoidable
354 * with shared fd spaces, we cannot solve it inside kernel,
355 * but we take care of internal coherence yet.
356 */
357
7cbe66b6 358static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
1da177e4 359{
7cbe66b6 360 struct qstr name = { .name = "" };
2c48b9c4 361 struct path path;
7cbe66b6 362 struct file *file;
1da177e4 363 int fd;
1da177e4 364
a677a039 365 fd = get_unused_fd_flags(flags);
7cbe66b6
AV
366 if (unlikely(fd < 0))
367 return fd;
1da177e4 368
4b936885 369 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
2c48b9c4 370 if (unlikely(!path.dentry)) {
7cbe66b6 371 put_unused_fd(fd);
39d8c1b6 372 return -ENOMEM;
7cbe66b6 373 }
2c48b9c4 374 path.mnt = mntget(sock_mnt);
39d8c1b6 375
2c48b9c4 376 d_instantiate(path.dentry, SOCK_INODE(sock));
cc3808f8 377 SOCK_INODE(sock)->i_fop = &socket_file_ops;
39d8c1b6 378
2c48b9c4 379 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
ce8d2cdf 380 &socket_file_ops);
cc3808f8
AV
381 if (unlikely(!file)) {
382 /* drop dentry, keep inode */
7de9c6ee 383 ihold(path.dentry->d_inode);
2c48b9c4 384 path_put(&path);
cc3808f8
AV
385 put_unused_fd(fd);
386 return -ENFILE;
387 }
388
389 sock->file = file;
77d27200 390 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
39d8c1b6
DM
391 file->f_pos = 0;
392 file->private_data = sock;
1da177e4 393
7cbe66b6
AV
394 *f = file;
395 return fd;
39d8c1b6
DM
396}
397
a677a039 398int sock_map_fd(struct socket *sock, int flags)
39d8c1b6
DM
399{
400 struct file *newfile;
7cbe66b6 401 int fd = sock_alloc_file(sock, &newfile, flags);
39d8c1b6 402
7cbe66b6 403 if (likely(fd >= 0))
39d8c1b6 404 fd_install(fd, newfile);
7cbe66b6 405
1da177e4
LT
406 return fd;
407}
c6d409cf 408EXPORT_SYMBOL(sock_map_fd);
1da177e4 409
6cb153ca
BL
410static struct socket *sock_from_file(struct file *file, int *err)
411{
6cb153ca
BL
412 if (file->f_op == &socket_file_ops)
413 return file->private_data; /* set in sock_map_fd */
414
23bb80d2
ED
415 *err = -ENOTSOCK;
416 return NULL;
6cb153ca
BL
417}
418
1da177e4 419/**
c6d409cf 420 * sockfd_lookup - Go from a file number to its socket slot
1da177e4
LT
421 * @fd: file handle
422 * @err: pointer to an error code return
423 *
424 * The file handle passed in is locked and the socket it is bound
425 * too is returned. If an error occurs the err pointer is overwritten
426 * with a negative errno code and NULL is returned. The function checks
427 * for both invalid handles and passing a handle which is not a socket.
428 *
429 * On a success the socket object pointer is returned.
430 */
431
432struct socket *sockfd_lookup(int fd, int *err)
433{
434 struct file *file;
1da177e4
LT
435 struct socket *sock;
436
89bddce5
SH
437 file = fget(fd);
438 if (!file) {
1da177e4
LT
439 *err = -EBADF;
440 return NULL;
441 }
89bddce5 442
6cb153ca
BL
443 sock = sock_from_file(file, err);
444 if (!sock)
1da177e4 445 fput(file);
6cb153ca
BL
446 return sock;
447}
c6d409cf 448EXPORT_SYMBOL(sockfd_lookup);
1da177e4 449
6cb153ca
BL
450static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
451{
452 struct file *file;
453 struct socket *sock;
454
3672558c 455 *err = -EBADF;
6cb153ca
BL
456 file = fget_light(fd, fput_needed);
457 if (file) {
458 sock = sock_from_file(file, err);
459 if (sock)
460 return sock;
461 fput_light(file, *fput_needed);
1da177e4 462 }
6cb153ca 463 return NULL;
1da177e4
LT
464}
465
466/**
467 * sock_alloc - allocate a socket
89bddce5 468 *
1da177e4
LT
469 * Allocate a new inode and socket object. The two are bound together
470 * and initialised. The socket is then returned. If we are out of inodes
471 * NULL is returned.
472 */
473
474static struct socket *sock_alloc(void)
475{
89bddce5
SH
476 struct inode *inode;
477 struct socket *sock;
1da177e4
LT
478
479 inode = new_inode(sock_mnt->mnt_sb);
480 if (!inode)
481 return NULL;
482
483 sock = SOCKET_I(inode);
484
29a020d3 485 kmemcheck_annotate_bitfield(sock, type);
85fe4025 486 inode->i_ino = get_next_ino();
89bddce5 487 inode->i_mode = S_IFSOCK | S_IRWXUGO;
8192b0c4
DH
488 inode->i_uid = current_fsuid();
489 inode->i_gid = current_fsgid();
1da177e4 490
4e69489a 491 percpu_add(sockets_in_use, 1);
1da177e4
LT
492 return sock;
493}
494
495/*
496 * In theory you can't get an open on this inode, but /proc provides
497 * a back door. Remember to keep it shut otherwise you'll let the
498 * creepy crawlies in.
499 */
89bddce5 500
1da177e4
LT
501static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
502{
503 return -ENXIO;
504}
505
4b6f5d20 506const struct file_operations bad_sock_fops = {
1da177e4
LT
507 .owner = THIS_MODULE,
508 .open = sock_no_open,
6038f373 509 .llseek = noop_llseek,
1da177e4
LT
510};
511
512/**
513 * sock_release - close a socket
514 * @sock: socket to close
515 *
516 * The socket is released from the protocol stack if it has a release
517 * callback, and the inode is then released if the socket is bound to
89bddce5 518 * an inode not a file.
1da177e4 519 */
89bddce5 520
1da177e4
LT
521void sock_release(struct socket *sock)
522{
523 if (sock->ops) {
524 struct module *owner = sock->ops->owner;
525
526 sock->ops->release(sock);
527 sock->ops = NULL;
528 module_put(owner);
529 }
530
eaefd110 531 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
1da177e4
LT
532 printk(KERN_ERR "sock_release: fasync list not empty!\n");
533
4e69489a 534 percpu_sub(sockets_in_use, 1);
1da177e4
LT
535 if (!sock->file) {
536 iput(SOCK_INODE(sock));
537 return;
538 }
89bddce5 539 sock->file = NULL;
1da177e4 540}
c6d409cf 541EXPORT_SYMBOL(sock_release);
1da177e4 542
2244d07b 543int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
20d49473 544{
2244d07b 545 *tx_flags = 0;
20d49473 546 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
2244d07b 547 *tx_flags |= SKBTX_HW_TSTAMP;
20d49473 548 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
2244d07b 549 *tx_flags |= SKBTX_SW_TSTAMP;
20d49473
PO
550 return 0;
551}
552EXPORT_SYMBOL(sock_tx_timestamp);
553
89bddce5 554static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
1da177e4
LT
555 struct msghdr *msg, size_t size)
556{
557 struct sock_iocb *si = kiocb_to_siocb(iocb);
558 int err;
559
f8451725
HX
560 sock_update_classid(sock->sk);
561
1da177e4
LT
562 si->sock = sock;
563 si->scm = NULL;
564 si->msg = msg;
565 si->size = size;
566
567 err = security_socket_sendmsg(sock, msg, size);
568 if (err)
569 return err;
570
571 return sock->ops->sendmsg(iocb, sock, msg, size);
572}
573
574int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
575{
576 struct kiocb iocb;
577 struct sock_iocb siocb;
578 int ret;
579
580 init_sync_kiocb(&iocb, NULL);
581 iocb.private = &siocb;
582 ret = __sock_sendmsg(&iocb, sock, msg, size);
583 if (-EIOCBQUEUED == ret)
584 ret = wait_on_sync_kiocb(&iocb);
585 return ret;
586}
c6d409cf 587EXPORT_SYMBOL(sock_sendmsg);
1da177e4
LT
588
589int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
590 struct kvec *vec, size_t num, size_t size)
591{
592 mm_segment_t oldfs = get_fs();
593 int result;
594
595 set_fs(KERNEL_DS);
596 /*
597 * the following is safe, since for compiler definitions of kvec and
598 * iovec are identical, yielding the same in-core layout and alignment
599 */
89bddce5 600 msg->msg_iov = (struct iovec *)vec;
1da177e4
LT
601 msg->msg_iovlen = num;
602 result = sock_sendmsg(sock, msg, size);
603 set_fs(oldfs);
604 return result;
605}
c6d409cf 606EXPORT_SYMBOL(kernel_sendmsg);
1da177e4 607
20d49473
PO
608static int ktime2ts(ktime_t kt, struct timespec *ts)
609{
610 if (kt.tv64) {
611 *ts = ktime_to_timespec(kt);
612 return 1;
613 } else {
614 return 0;
615 }
616}
617
92f37fd2
ED
618/*
619 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
620 */
621void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
622 struct sk_buff *skb)
623{
20d49473
PO
624 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
625 struct timespec ts[3];
626 int empty = 1;
627 struct skb_shared_hwtstamps *shhwtstamps =
628 skb_hwtstamps(skb);
629
630 /* Race occurred between timestamp enabling and packet
631 receiving. Fill in the current time for now. */
632 if (need_software_tstamp && skb->tstamp.tv64 == 0)
633 __net_timestamp(skb);
634
635 if (need_software_tstamp) {
636 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
637 struct timeval tv;
638 skb_get_timestamp(skb, &tv);
639 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
640 sizeof(tv), &tv);
641 } else {
842509b8 642 skb_get_timestampns(skb, &ts[0]);
20d49473 643 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
842509b8 644 sizeof(ts[0]), &ts[0]);
20d49473
PO
645 }
646 }
647
648
649 memset(ts, 0, sizeof(ts));
650 if (skb->tstamp.tv64 &&
651 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
652 skb_get_timestampns(skb, ts + 0);
653 empty = 0;
654 }
655 if (shhwtstamps) {
656 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
657 ktime2ts(shhwtstamps->syststamp, ts + 1))
658 empty = 0;
659 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
660 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
661 empty = 0;
92f37fd2 662 }
20d49473
PO
663 if (!empty)
664 put_cmsg(msg, SOL_SOCKET,
665 SCM_TIMESTAMPING, sizeof(ts), &ts);
92f37fd2 666}
7c81fd8b
ACM
667EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
668
11165f14 669static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
670 struct sk_buff *skb)
3b885787
NH
671{
672 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
673 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
674 sizeof(__u32), &skb->dropcount);
675}
676
767dd033 677void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
3b885787
NH
678 struct sk_buff *skb)
679{
680 sock_recv_timestamp(msg, sk, skb);
681 sock_recv_drops(msg, sk, skb);
682}
767dd033 683EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
3b885787 684
a2e27255
ACM
685static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
686 struct msghdr *msg, size_t size, int flags)
1da177e4 687{
1da177e4
LT
688 struct sock_iocb *si = kiocb_to_siocb(iocb);
689
f8451725
HX
690 sock_update_classid(sock->sk);
691
1da177e4
LT
692 si->sock = sock;
693 si->scm = NULL;
694 si->msg = msg;
695 si->size = size;
696 si->flags = flags;
697
1da177e4
LT
698 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
699}
700
a2e27255
ACM
701static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
702 struct msghdr *msg, size_t size, int flags)
703{
704 int err = security_socket_recvmsg(sock, msg, size, flags);
705
706 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
707}
708
89bddce5 709int sock_recvmsg(struct socket *sock, struct msghdr *msg,
1da177e4
LT
710 size_t size, int flags)
711{
712 struct kiocb iocb;
713 struct sock_iocb siocb;
714 int ret;
715
89bddce5 716 init_sync_kiocb(&iocb, NULL);
1da177e4
LT
717 iocb.private = &siocb;
718 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
719 if (-EIOCBQUEUED == ret)
720 ret = wait_on_sync_kiocb(&iocb);
721 return ret;
722}
c6d409cf 723EXPORT_SYMBOL(sock_recvmsg);
1da177e4 724
a2e27255
ACM
725static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
726 size_t size, int flags)
727{
728 struct kiocb iocb;
729 struct sock_iocb siocb;
730 int ret;
731
732 init_sync_kiocb(&iocb, NULL);
733 iocb.private = &siocb;
734 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
735 if (-EIOCBQUEUED == ret)
736 ret = wait_on_sync_kiocb(&iocb);
737 return ret;
738}
739
c1249c0a
ML
740/**
741 * kernel_recvmsg - Receive a message from a socket (kernel space)
742 * @sock: The socket to receive the message from
743 * @msg: Received message
744 * @vec: Input s/g array for message data
745 * @num: Size of input s/g array
746 * @size: Number of bytes to read
747 * @flags: Message flags (MSG_DONTWAIT, etc...)
748 *
749 * On return the msg structure contains the scatter/gather array passed in the
750 * vec argument. The array is modified so that it consists of the unfilled
751 * portion of the original array.
752 *
753 * The returned value is the total number of bytes received, or an error.
754 */
89bddce5
SH
755int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
756 struct kvec *vec, size_t num, size_t size, int flags)
1da177e4
LT
757{
758 mm_segment_t oldfs = get_fs();
759 int result;
760
761 set_fs(KERNEL_DS);
762 /*
763 * the following is safe, since for compiler definitions of kvec and
764 * iovec are identical, yielding the same in-core layout and alignment
765 */
89bddce5 766 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
1da177e4
LT
767 result = sock_recvmsg(sock, msg, size, flags);
768 set_fs(oldfs);
769 return result;
770}
c6d409cf 771EXPORT_SYMBOL(kernel_recvmsg);
1da177e4
LT
772
773static void sock_aio_dtor(struct kiocb *iocb)
774{
775 kfree(iocb->private);
776}
777
ce1d4d3e
CH
778static ssize_t sock_sendpage(struct file *file, struct page *page,
779 int offset, size_t size, loff_t *ppos, int more)
1da177e4 780{
1da177e4
LT
781 struct socket *sock;
782 int flags;
783
ce1d4d3e
CH
784 sock = file->private_data;
785
786 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
787 if (more)
788 flags |= MSG_MORE;
789
e6949583 790 return kernel_sendpage(sock, page, offset, size, flags);
ce1d4d3e 791}
1da177e4 792
9c55e01c 793static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
c6d409cf 794 struct pipe_inode_info *pipe, size_t len,
9c55e01c
JA
795 unsigned int flags)
796{
797 struct socket *sock = file->private_data;
798
997b37da
RDC
799 if (unlikely(!sock->ops->splice_read))
800 return -EINVAL;
801
f8451725
HX
802 sock_update_classid(sock->sk);
803
9c55e01c
JA
804 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
805}
806
ce1d4d3e 807static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
89bddce5 808 struct sock_iocb *siocb)
ce1d4d3e
CH
809{
810 if (!is_sync_kiocb(iocb)) {
811 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
812 if (!siocb)
813 return NULL;
1da177e4
LT
814 iocb->ki_dtor = sock_aio_dtor;
815 }
1da177e4 816
ce1d4d3e 817 siocb->kiocb = iocb;
ce1d4d3e
CH
818 iocb->private = siocb;
819 return siocb;
1da177e4
LT
820}
821
ce1d4d3e 822static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
027445c3
BP
823 struct file *file, const struct iovec *iov,
824 unsigned long nr_segs)
ce1d4d3e
CH
825{
826 struct socket *sock = file->private_data;
827 size_t size = 0;
828 int i;
1da177e4 829
89bddce5
SH
830 for (i = 0; i < nr_segs; i++)
831 size += iov[i].iov_len;
1da177e4 832
ce1d4d3e
CH
833 msg->msg_name = NULL;
834 msg->msg_namelen = 0;
835 msg->msg_control = NULL;
836 msg->msg_controllen = 0;
89bddce5 837 msg->msg_iov = (struct iovec *)iov;
ce1d4d3e
CH
838 msg->msg_iovlen = nr_segs;
839 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
840
841 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
842}
843
027445c3
BP
844static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
845 unsigned long nr_segs, loff_t pos)
ce1d4d3e
CH
846{
847 struct sock_iocb siocb, *x;
848
1da177e4
LT
849 if (pos != 0)
850 return -ESPIPE;
027445c3
BP
851
852 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
1da177e4
LT
853 return 0;
854
027445c3
BP
855
856 x = alloc_sock_iocb(iocb, &siocb);
ce1d4d3e
CH
857 if (!x)
858 return -ENOMEM;
027445c3 859 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
1da177e4
LT
860}
861
ce1d4d3e 862static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
027445c3
BP
863 struct file *file, const struct iovec *iov,
864 unsigned long nr_segs)
1da177e4 865{
ce1d4d3e
CH
866 struct socket *sock = file->private_data;
867 size_t size = 0;
868 int i;
1da177e4 869
89bddce5
SH
870 for (i = 0; i < nr_segs; i++)
871 size += iov[i].iov_len;
1da177e4 872
ce1d4d3e
CH
873 msg->msg_name = NULL;
874 msg->msg_namelen = 0;
875 msg->msg_control = NULL;
876 msg->msg_controllen = 0;
89bddce5 877 msg->msg_iov = (struct iovec *)iov;
ce1d4d3e
CH
878 msg->msg_iovlen = nr_segs;
879 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
880 if (sock->type == SOCK_SEQPACKET)
881 msg->msg_flags |= MSG_EOR;
1da177e4 882
ce1d4d3e 883 return __sock_sendmsg(iocb, sock, msg, size);
1da177e4
LT
884}
885
027445c3
BP
886static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
887 unsigned long nr_segs, loff_t pos)
ce1d4d3e
CH
888{
889 struct sock_iocb siocb, *x;
1da177e4 890
ce1d4d3e
CH
891 if (pos != 0)
892 return -ESPIPE;
027445c3 893
027445c3 894 x = alloc_sock_iocb(iocb, &siocb);
ce1d4d3e
CH
895 if (!x)
896 return -ENOMEM;
1da177e4 897
027445c3 898 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
1da177e4
LT
899}
900
1da177e4
LT
901/*
902 * Atomic setting of ioctl hooks to avoid race
903 * with module unload.
904 */
905
4a3e2f71 906static DEFINE_MUTEX(br_ioctl_mutex);
c6d409cf 907static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1da177e4 908
881d966b 909void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1da177e4 910{
4a3e2f71 911 mutex_lock(&br_ioctl_mutex);
1da177e4 912 br_ioctl_hook = hook;
4a3e2f71 913 mutex_unlock(&br_ioctl_mutex);
1da177e4
LT
914}
915EXPORT_SYMBOL(brioctl_set);
916
4a3e2f71 917static DEFINE_MUTEX(vlan_ioctl_mutex);
881d966b 918static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1da177e4 919
881d966b 920void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1da177e4 921{
4a3e2f71 922 mutex_lock(&vlan_ioctl_mutex);
1da177e4 923 vlan_ioctl_hook = hook;
4a3e2f71 924 mutex_unlock(&vlan_ioctl_mutex);
1da177e4
LT
925}
926EXPORT_SYMBOL(vlan_ioctl_set);
927
4a3e2f71 928static DEFINE_MUTEX(dlci_ioctl_mutex);
89bddce5 929static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1da177e4 930
89bddce5 931void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1da177e4 932{
4a3e2f71 933 mutex_lock(&dlci_ioctl_mutex);
1da177e4 934 dlci_ioctl_hook = hook;
4a3e2f71 935 mutex_unlock(&dlci_ioctl_mutex);
1da177e4
LT
936}
937EXPORT_SYMBOL(dlci_ioctl_set);
938
6b96018b
AB
939static long sock_do_ioctl(struct net *net, struct socket *sock,
940 unsigned int cmd, unsigned long arg)
941{
942 int err;
943 void __user *argp = (void __user *)arg;
944
945 err = sock->ops->ioctl(sock, cmd, arg);
946
947 /*
948 * If this ioctl is unknown try to hand it down
949 * to the NIC driver.
950 */
951 if (err == -ENOIOCTLCMD)
952 err = dev_ioctl(net, cmd, argp);
953
954 return err;
955}
956
1da177e4
LT
957/*
958 * With an ioctl, arg may well be a user mode pointer, but we don't know
959 * what to do with it - that's up to the protocol still.
960 */
961
962static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
963{
964 struct socket *sock;
881d966b 965 struct sock *sk;
1da177e4
LT
966 void __user *argp = (void __user *)arg;
967 int pid, err;
881d966b 968 struct net *net;
1da177e4 969
b69aee04 970 sock = file->private_data;
881d966b 971 sk = sock->sk;
3b1e0a65 972 net = sock_net(sk);
1da177e4 973 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
881d966b 974 err = dev_ioctl(net, cmd, argp);
1da177e4 975 } else
3d23e349 976#ifdef CONFIG_WEXT_CORE
1da177e4 977 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
881d966b 978 err = dev_ioctl(net, cmd, argp);
1da177e4 979 } else
3d23e349 980#endif
89bddce5 981 switch (cmd) {
1da177e4
LT
982 case FIOSETOWN:
983 case SIOCSPGRP:
984 err = -EFAULT;
985 if (get_user(pid, (int __user *)argp))
986 break;
987 err = f_setown(sock->file, pid, 1);
988 break;
989 case FIOGETOWN:
990 case SIOCGPGRP:
609d7fa9 991 err = put_user(f_getown(sock->file),
89bddce5 992 (int __user *)argp);
1da177e4
LT
993 break;
994 case SIOCGIFBR:
995 case SIOCSIFBR:
996 case SIOCBRADDBR:
997 case SIOCBRDELBR:
998 err = -ENOPKG;
999 if (!br_ioctl_hook)
1000 request_module("bridge");
1001
4a3e2f71 1002 mutex_lock(&br_ioctl_mutex);
89bddce5 1003 if (br_ioctl_hook)
881d966b 1004 err = br_ioctl_hook(net, cmd, argp);
4a3e2f71 1005 mutex_unlock(&br_ioctl_mutex);
1da177e4
LT
1006 break;
1007 case SIOCGIFVLAN:
1008 case SIOCSIFVLAN:
1009 err = -ENOPKG;
1010 if (!vlan_ioctl_hook)
1011 request_module("8021q");
1012
4a3e2f71 1013 mutex_lock(&vlan_ioctl_mutex);
1da177e4 1014 if (vlan_ioctl_hook)
881d966b 1015 err = vlan_ioctl_hook(net, argp);
4a3e2f71 1016 mutex_unlock(&vlan_ioctl_mutex);
1da177e4 1017 break;
1da177e4
LT
1018 case SIOCADDDLCI:
1019 case SIOCDELDLCI:
1020 err = -ENOPKG;
1021 if (!dlci_ioctl_hook)
1022 request_module("dlci");
1023
7512cbf6
PE
1024 mutex_lock(&dlci_ioctl_mutex);
1025 if (dlci_ioctl_hook)
1da177e4 1026 err = dlci_ioctl_hook(cmd, argp);
7512cbf6 1027 mutex_unlock(&dlci_ioctl_mutex);
1da177e4
LT
1028 break;
1029 default:
6b96018b 1030 err = sock_do_ioctl(net, sock, cmd, arg);
1da177e4 1031 break;
89bddce5 1032 }
1da177e4
LT
1033 return err;
1034}
1035
1036int sock_create_lite(int family, int type, int protocol, struct socket **res)
1037{
1038 int err;
1039 struct socket *sock = NULL;
89bddce5 1040
1da177e4
LT
1041 err = security_socket_create(family, type, protocol, 1);
1042 if (err)
1043 goto out;
1044
1045 sock = sock_alloc();
1046 if (!sock) {
1047 err = -ENOMEM;
1048 goto out;
1049 }
1050
1da177e4 1051 sock->type = type;
7420ed23
VY
1052 err = security_socket_post_create(sock, family, type, protocol, 1);
1053 if (err)
1054 goto out_release;
1055
1da177e4
LT
1056out:
1057 *res = sock;
1058 return err;
7420ed23
VY
1059out_release:
1060 sock_release(sock);
1061 sock = NULL;
1062 goto out;
1da177e4 1063}
c6d409cf 1064EXPORT_SYMBOL(sock_create_lite);
1da177e4
LT
1065
1066/* No kernel lock held - perfect */
89bddce5 1067static unsigned int sock_poll(struct file *file, poll_table *wait)
1da177e4
LT
1068{
1069 struct socket *sock;
1070
1071 /*
89bddce5 1072 * We can't return errors to poll, so it's either yes or no.
1da177e4 1073 */
b69aee04 1074 sock = file->private_data;
1da177e4
LT
1075 return sock->ops->poll(file, sock, wait);
1076}
1077
89bddce5 1078static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1da177e4 1079{
b69aee04 1080 struct socket *sock = file->private_data;
1da177e4
LT
1081
1082 return sock->ops->mmap(file, sock, vma);
1083}
1084
20380731 1085static int sock_close(struct inode *inode, struct file *filp)
1da177e4
LT
1086{
1087 /*
89bddce5
SH
1088 * It was possible the inode is NULL we were
1089 * closing an unfinished socket.
1da177e4
LT
1090 */
1091
89bddce5 1092 if (!inode) {
1da177e4
LT
1093 printk(KERN_DEBUG "sock_close: NULL inode\n");
1094 return 0;
1095 }
1da177e4
LT
1096 sock_release(SOCKET_I(inode));
1097 return 0;
1098}
1099
1100/*
1101 * Update the socket async list
1102 *
1103 * Fasync_list locking strategy.
1104 *
1105 * 1. fasync_list is modified only under process context socket lock
1106 * i.e. under semaphore.
1107 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
989a2979 1108 * or under socket lock
1da177e4
LT
1109 */
1110
1111static int sock_fasync(int fd, struct file *filp, int on)
1112{
989a2979
ED
1113 struct socket *sock = filp->private_data;
1114 struct sock *sk = sock->sk;
eaefd110 1115 struct socket_wq *wq;
1da177e4 1116
989a2979 1117 if (sk == NULL)
1da177e4 1118 return -EINVAL;
1da177e4
LT
1119
1120 lock_sock(sk);
eaefd110
ED
1121 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1122 fasync_helper(fd, filp, on, &wq->fasync_list);
1da177e4 1123
eaefd110 1124 if (!wq->fasync_list)
989a2979
ED
1125 sock_reset_flag(sk, SOCK_FASYNC);
1126 else
bcdce719 1127 sock_set_flag(sk, SOCK_FASYNC);
1da177e4 1128
989a2979 1129 release_sock(sk);
1da177e4
LT
1130 return 0;
1131}
1132
43815482 1133/* This function may be called only under socket lock or callback_lock or rcu_lock */
1da177e4
LT
1134
1135int sock_wake_async(struct socket *sock, int how, int band)
1136{
43815482
ED
1137 struct socket_wq *wq;
1138
1139 if (!sock)
1140 return -1;
1141 rcu_read_lock();
1142 wq = rcu_dereference(sock->wq);
1143 if (!wq || !wq->fasync_list) {
1144 rcu_read_unlock();
1da177e4 1145 return -1;
43815482 1146 }
89bddce5 1147 switch (how) {
8d8ad9d7 1148 case SOCK_WAKE_WAITD:
1da177e4
LT
1149 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1150 break;
1151 goto call_kill;
8d8ad9d7 1152 case SOCK_WAKE_SPACE:
1da177e4
LT
1153 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1154 break;
1155 /* fall through */
8d8ad9d7 1156 case SOCK_WAKE_IO:
89bddce5 1157call_kill:
43815482 1158 kill_fasync(&wq->fasync_list, SIGIO, band);
1da177e4 1159 break;
8d8ad9d7 1160 case SOCK_WAKE_URG:
43815482 1161 kill_fasync(&wq->fasync_list, SIGURG, band);
1da177e4 1162 }
43815482 1163 rcu_read_unlock();
1da177e4
LT
1164 return 0;
1165}
c6d409cf 1166EXPORT_SYMBOL(sock_wake_async);
1da177e4 1167
721db93a 1168int __sock_create(struct net *net, int family, int type, int protocol,
89bddce5 1169 struct socket **res, int kern)
1da177e4
LT
1170{
1171 int err;
1172 struct socket *sock;
55737fda 1173 const struct net_proto_family *pf;
1da177e4
LT
1174
1175 /*
89bddce5 1176 * Check protocol is in range
1da177e4
LT
1177 */
1178 if (family < 0 || family >= NPROTO)
1179 return -EAFNOSUPPORT;
1180 if (type < 0 || type >= SOCK_MAX)
1181 return -EINVAL;
1182
1183 /* Compatibility.
1184
1185 This uglymoron is moved from INET layer to here to avoid
1186 deadlock in module load.
1187 */
1188 if (family == PF_INET && type == SOCK_PACKET) {
89bddce5 1189 static int warned;
1da177e4
LT
1190 if (!warned) {
1191 warned = 1;
89bddce5
SH
1192 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1193 current->comm);
1da177e4
LT
1194 }
1195 family = PF_PACKET;
1196 }
1197
1198 err = security_socket_create(family, type, protocol, kern);
1199 if (err)
1200 return err;
89bddce5 1201
55737fda
SH
1202 /*
1203 * Allocate the socket and allow the family to set things up. if
1204 * the protocol is 0, the family is instructed to select an appropriate
1205 * default.
1206 */
1207 sock = sock_alloc();
1208 if (!sock) {
1209 if (net_ratelimit())
1210 printk(KERN_WARNING "socket: no more sockets\n");
1211 return -ENFILE; /* Not exactly a match, but its the
1212 closest posix thing */
1213 }
1214
1215 sock->type = type;
1216
95a5afca 1217#ifdef CONFIG_MODULES
89bddce5
SH
1218 /* Attempt to load a protocol module if the find failed.
1219 *
1220 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1da177e4
LT
1221 * requested real, full-featured networking support upon configuration.
1222 * Otherwise module support will break!
1223 */
190683a9 1224 if (rcu_access_pointer(net_families[family]) == NULL)
89bddce5 1225 request_module("net-pf-%d", family);
1da177e4
LT
1226#endif
1227
55737fda
SH
1228 rcu_read_lock();
1229 pf = rcu_dereference(net_families[family]);
1230 err = -EAFNOSUPPORT;
1231 if (!pf)
1232 goto out_release;
1da177e4
LT
1233
1234 /*
1235 * We will call the ->create function, that possibly is in a loadable
1236 * module, so we have to bump that loadable module refcnt first.
1237 */
55737fda 1238 if (!try_module_get(pf->owner))
1da177e4
LT
1239 goto out_release;
1240
55737fda
SH
1241 /* Now protected by module ref count */
1242 rcu_read_unlock();
1243
3f378b68 1244 err = pf->create(net, sock, protocol, kern);
55737fda 1245 if (err < 0)
1da177e4 1246 goto out_module_put;
a79af59e 1247
1da177e4
LT
1248 /*
1249 * Now to bump the refcnt of the [loadable] module that owns this
1250 * socket at sock_release time we decrement its refcnt.
1251 */
55737fda
SH
1252 if (!try_module_get(sock->ops->owner))
1253 goto out_module_busy;
1254
1da177e4
LT
1255 /*
1256 * Now that we're done with the ->create function, the [loadable]
1257 * module can have its refcnt decremented
1258 */
55737fda 1259 module_put(pf->owner);
7420ed23
VY
1260 err = security_socket_post_create(sock, family, type, protocol, kern);
1261 if (err)
3b185525 1262 goto out_sock_release;
55737fda 1263 *res = sock;
1da177e4 1264
55737fda
SH
1265 return 0;
1266
1267out_module_busy:
1268 err = -EAFNOSUPPORT;
1da177e4 1269out_module_put:
55737fda
SH
1270 sock->ops = NULL;
1271 module_put(pf->owner);
1272out_sock_release:
1da177e4 1273 sock_release(sock);
55737fda
SH
1274 return err;
1275
1276out_release:
1277 rcu_read_unlock();
1278 goto out_sock_release;
1da177e4 1279}
721db93a 1280EXPORT_SYMBOL(__sock_create);
1da177e4
LT
1281
1282int sock_create(int family, int type, int protocol, struct socket **res)
1283{
1b8d7ae4 1284 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1da177e4 1285}
c6d409cf 1286EXPORT_SYMBOL(sock_create);
1da177e4
LT
1287
1288int sock_create_kern(int family, int type, int protocol, struct socket **res)
1289{
1b8d7ae4 1290 return __sock_create(&init_net, family, type, protocol, res, 1);
1da177e4 1291}
c6d409cf 1292EXPORT_SYMBOL(sock_create_kern);
1da177e4 1293
3e0fa65f 1294SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1da177e4
LT
1295{
1296 int retval;
1297 struct socket *sock;
a677a039
UD
1298 int flags;
1299
e38b36f3
UD
1300 /* Check the SOCK_* constants for consistency. */
1301 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1302 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1303 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1304 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1305
a677a039 1306 flags = type & ~SOCK_TYPE_MASK;
77d27200 1307 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
a677a039
UD
1308 return -EINVAL;
1309 type &= SOCK_TYPE_MASK;
1da177e4 1310
aaca0bdc
UD
1311 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1312 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1313
1da177e4
LT
1314 retval = sock_create(family, type, protocol, &sock);
1315 if (retval < 0)
1316 goto out;
1317
77d27200 1318 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1da177e4
LT
1319 if (retval < 0)
1320 goto out_release;
1321
1322out:
1323 /* It may be already another descriptor 8) Not kernel problem. */
1324 return retval;
1325
1326out_release:
1327 sock_release(sock);
1328 return retval;
1329}
1330
1331/*
1332 * Create a pair of connected sockets.
1333 */
1334
3e0fa65f
HC
1335SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1336 int __user *, usockvec)
1da177e4
LT
1337{
1338 struct socket *sock1, *sock2;
1339 int fd1, fd2, err;
db349509 1340 struct file *newfile1, *newfile2;
a677a039
UD
1341 int flags;
1342
1343 flags = type & ~SOCK_TYPE_MASK;
77d27200 1344 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
a677a039
UD
1345 return -EINVAL;
1346 type &= SOCK_TYPE_MASK;
1da177e4 1347
aaca0bdc
UD
1348 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1349 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1350
1da177e4
LT
1351 /*
1352 * Obtain the first socket and check if the underlying protocol
1353 * supports the socketpair call.
1354 */
1355
1356 err = sock_create(family, type, protocol, &sock1);
1357 if (err < 0)
1358 goto out;
1359
1360 err = sock_create(family, type, protocol, &sock2);
1361 if (err < 0)
1362 goto out_release_1;
1363
1364 err = sock1->ops->socketpair(sock1, sock2);
89bddce5 1365 if (err < 0)
1da177e4
LT
1366 goto out_release_both;
1367
7cbe66b6 1368 fd1 = sock_alloc_file(sock1, &newfile1, flags);
bf3c23d1
DM
1369 if (unlikely(fd1 < 0)) {
1370 err = fd1;
db349509 1371 goto out_release_both;
bf3c23d1 1372 }
1da177e4 1373
7cbe66b6 1374 fd2 = sock_alloc_file(sock2, &newfile2, flags);
198de4d7
AV
1375 if (unlikely(fd2 < 0)) {
1376 err = fd2;
1377 fput(newfile1);
1378 put_unused_fd(fd1);
1379 sock_release(sock2);
1380 goto out;
db349509
AV
1381 }
1382
157cf649 1383 audit_fd_pair(fd1, fd2);
db349509
AV
1384 fd_install(fd1, newfile1);
1385 fd_install(fd2, newfile2);
1da177e4
LT
1386 /* fd1 and fd2 may be already another descriptors.
1387 * Not kernel problem.
1388 */
1389
89bddce5 1390 err = put_user(fd1, &usockvec[0]);
1da177e4
LT
1391 if (!err)
1392 err = put_user(fd2, &usockvec[1]);
1393 if (!err)
1394 return 0;
1395
1396 sys_close(fd2);
1397 sys_close(fd1);
1398 return err;
1399
1da177e4 1400out_release_both:
89bddce5 1401 sock_release(sock2);
1da177e4 1402out_release_1:
89bddce5 1403 sock_release(sock1);
1da177e4
LT
1404out:
1405 return err;
1406}
1407
1da177e4
LT
1408/*
1409 * Bind a name to a socket. Nothing much to do here since it's
1410 * the protocol's responsibility to handle the local address.
1411 *
1412 * We move the socket address to kernel space before we call
1413 * the protocol layer (having also checked the address is ok).
1414 */
1415
20f37034 1416SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1da177e4
LT
1417{
1418 struct socket *sock;
230b1839 1419 struct sockaddr_storage address;
6cb153ca 1420 int err, fput_needed;
1da177e4 1421
89bddce5 1422 sock = sockfd_lookup_light(fd, &err, &fput_needed);
e71a4783 1423 if (sock) {
230b1839 1424 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
89bddce5
SH
1425 if (err >= 0) {
1426 err = security_socket_bind(sock,
230b1839 1427 (struct sockaddr *)&address,
89bddce5 1428 addrlen);
6cb153ca
BL
1429 if (!err)
1430 err = sock->ops->bind(sock,
89bddce5 1431 (struct sockaddr *)
230b1839 1432 &address, addrlen);
1da177e4 1433 }
6cb153ca 1434 fput_light(sock->file, fput_needed);
89bddce5 1435 }
1da177e4
LT
1436 return err;
1437}
1438
1da177e4
LT
1439/*
1440 * Perform a listen. Basically, we allow the protocol to do anything
1441 * necessary for a listen, and if that works, we mark the socket as
1442 * ready for listening.
1443 */
1444
3e0fa65f 1445SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1da177e4
LT
1446{
1447 struct socket *sock;
6cb153ca 1448 int err, fput_needed;
b8e1f9b5 1449 int somaxconn;
89bddce5
SH
1450
1451 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1452 if (sock) {
8efa6e93 1453 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
b8e1f9b5
PE
1454 if ((unsigned)backlog > somaxconn)
1455 backlog = somaxconn;
1da177e4
LT
1456
1457 err = security_socket_listen(sock, backlog);
6cb153ca
BL
1458 if (!err)
1459 err = sock->ops->listen(sock, backlog);
1da177e4 1460
6cb153ca 1461 fput_light(sock->file, fput_needed);
1da177e4
LT
1462 }
1463 return err;
1464}
1465
1da177e4
LT
1466/*
1467 * For accept, we attempt to create a new socket, set up the link
1468 * with the client, wake up the client, then return the new
1469 * connected fd. We collect the address of the connector in kernel
1470 * space and move it to user at the very end. This is unclean because
1471 * we open the socket then return an error.
1472 *
1473 * 1003.1g adds the ability to recvmsg() to query connection pending
1474 * status to recvmsg. We need to add that support in a way thats
1475 * clean when we restucture accept also.
1476 */
1477
20f37034
HC
1478SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1479 int __user *, upeer_addrlen, int, flags)
1da177e4
LT
1480{
1481 struct socket *sock, *newsock;
39d8c1b6 1482 struct file *newfile;
6cb153ca 1483 int err, len, newfd, fput_needed;
230b1839 1484 struct sockaddr_storage address;
1da177e4 1485
77d27200 1486 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
aaca0bdc
UD
1487 return -EINVAL;
1488
1489 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1490 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1491
6cb153ca 1492 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1493 if (!sock)
1494 goto out;
1495
1496 err = -ENFILE;
c6d409cf
ED
1497 newsock = sock_alloc();
1498 if (!newsock)
1da177e4
LT
1499 goto out_put;
1500
1501 newsock->type = sock->type;
1502 newsock->ops = sock->ops;
1503
1da177e4
LT
1504 /*
1505 * We don't need try_module_get here, as the listening socket (sock)
1506 * has the protocol module (sock->ops->owner) held.
1507 */
1508 __module_get(newsock->ops->owner);
1509
7cbe66b6 1510 newfd = sock_alloc_file(newsock, &newfile, flags);
39d8c1b6
DM
1511 if (unlikely(newfd < 0)) {
1512 err = newfd;
9a1875e6
DM
1513 sock_release(newsock);
1514 goto out_put;
39d8c1b6
DM
1515 }
1516
a79af59e
FF
1517 err = security_socket_accept(sock, newsock);
1518 if (err)
39d8c1b6 1519 goto out_fd;
a79af59e 1520
1da177e4
LT
1521 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1522 if (err < 0)
39d8c1b6 1523 goto out_fd;
1da177e4
LT
1524
1525 if (upeer_sockaddr) {
230b1839 1526 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
89bddce5 1527 &len, 2) < 0) {
1da177e4 1528 err = -ECONNABORTED;
39d8c1b6 1529 goto out_fd;
1da177e4 1530 }
230b1839
YH
1531 err = move_addr_to_user((struct sockaddr *)&address,
1532 len, upeer_sockaddr, upeer_addrlen);
1da177e4 1533 if (err < 0)
39d8c1b6 1534 goto out_fd;
1da177e4
LT
1535 }
1536
1537 /* File flags are not inherited via accept() unlike another OSes. */
1538
39d8c1b6
DM
1539 fd_install(newfd, newfile);
1540 err = newfd;
1da177e4 1541
1da177e4 1542out_put:
6cb153ca 1543 fput_light(sock->file, fput_needed);
1da177e4
LT
1544out:
1545 return err;
39d8c1b6 1546out_fd:
9606a216 1547 fput(newfile);
39d8c1b6 1548 put_unused_fd(newfd);
1da177e4
LT
1549 goto out_put;
1550}
1551
20f37034
HC
1552SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1553 int __user *, upeer_addrlen)
aaca0bdc 1554{
de11defe 1555 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
aaca0bdc
UD
1556}
1557
1da177e4
LT
1558/*
1559 * Attempt to connect to a socket with the server address. The address
1560 * is in user space so we verify it is OK and move it to kernel space.
1561 *
1562 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1563 * break bindings
1564 *
1565 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1566 * other SEQPACKET protocols that take time to connect() as it doesn't
1567 * include the -EINPROGRESS status for such sockets.
1568 */
1569
20f37034
HC
1570SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1571 int, addrlen)
1da177e4
LT
1572{
1573 struct socket *sock;
230b1839 1574 struct sockaddr_storage address;
6cb153ca 1575 int err, fput_needed;
1da177e4 1576
6cb153ca 1577 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1578 if (!sock)
1579 goto out;
230b1839 1580 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1da177e4
LT
1581 if (err < 0)
1582 goto out_put;
1583
89bddce5 1584 err =
230b1839 1585 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1da177e4
LT
1586 if (err)
1587 goto out_put;
1588
230b1839 1589 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1da177e4
LT
1590 sock->file->f_flags);
1591out_put:
6cb153ca 1592 fput_light(sock->file, fput_needed);
1da177e4
LT
1593out:
1594 return err;
1595}
1596
1597/*
1598 * Get the local address ('name') of a socket object. Move the obtained
1599 * name to user space.
1600 */
1601
20f37034
HC
1602SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1603 int __user *, usockaddr_len)
1da177e4
LT
1604{
1605 struct socket *sock;
230b1839 1606 struct sockaddr_storage address;
6cb153ca 1607 int len, err, fput_needed;
89bddce5 1608
6cb153ca 1609 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1610 if (!sock)
1611 goto out;
1612
1613 err = security_socket_getsockname(sock);
1614 if (err)
1615 goto out_put;
1616
230b1839 1617 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1da177e4
LT
1618 if (err)
1619 goto out_put;
230b1839 1620 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1da177e4
LT
1621
1622out_put:
6cb153ca 1623 fput_light(sock->file, fput_needed);
1da177e4
LT
1624out:
1625 return err;
1626}
1627
1628/*
1629 * Get the remote address ('name') of a socket object. Move the obtained
1630 * name to user space.
1631 */
1632
20f37034
HC
1633SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1634 int __user *, usockaddr_len)
1da177e4
LT
1635{
1636 struct socket *sock;
230b1839 1637 struct sockaddr_storage address;
6cb153ca 1638 int len, err, fput_needed;
1da177e4 1639
89bddce5
SH
1640 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1641 if (sock != NULL) {
1da177e4
LT
1642 err = security_socket_getpeername(sock);
1643 if (err) {
6cb153ca 1644 fput_light(sock->file, fput_needed);
1da177e4
LT
1645 return err;
1646 }
1647
89bddce5 1648 err =
230b1839 1649 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
89bddce5 1650 1);
1da177e4 1651 if (!err)
230b1839 1652 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
89bddce5 1653 usockaddr_len);
6cb153ca 1654 fput_light(sock->file, fput_needed);
1da177e4
LT
1655 }
1656 return err;
1657}
1658
1659/*
1660 * Send a datagram to a given address. We move the address into kernel
1661 * space and check the user space data area is readable before invoking
1662 * the protocol.
1663 */
1664
3e0fa65f
HC
1665SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1666 unsigned, flags, struct sockaddr __user *, addr,
1667 int, addr_len)
1da177e4
LT
1668{
1669 struct socket *sock;
230b1839 1670 struct sockaddr_storage address;
1da177e4
LT
1671 int err;
1672 struct msghdr msg;
1673 struct iovec iov;
6cb153ca 1674 int fput_needed;
6cb153ca 1675
253eacc0
LT
1676 if (len > INT_MAX)
1677 len = INT_MAX;
de0fa95c
PE
1678 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1679 if (!sock)
4387ff75 1680 goto out;
6cb153ca 1681
89bddce5
SH
1682 iov.iov_base = buff;
1683 iov.iov_len = len;
1684 msg.msg_name = NULL;
1685 msg.msg_iov = &iov;
1686 msg.msg_iovlen = 1;
1687 msg.msg_control = NULL;
1688 msg.msg_controllen = 0;
1689 msg.msg_namelen = 0;
6cb153ca 1690 if (addr) {
230b1839 1691 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1da177e4
LT
1692 if (err < 0)
1693 goto out_put;
230b1839 1694 msg.msg_name = (struct sockaddr *)&address;
89bddce5 1695 msg.msg_namelen = addr_len;
1da177e4
LT
1696 }
1697 if (sock->file->f_flags & O_NONBLOCK)
1698 flags |= MSG_DONTWAIT;
1699 msg.msg_flags = flags;
1700 err = sock_sendmsg(sock, &msg, len);
1701
89bddce5 1702out_put:
de0fa95c 1703 fput_light(sock->file, fput_needed);
4387ff75 1704out:
1da177e4
LT
1705 return err;
1706}
1707
1708/*
89bddce5 1709 * Send a datagram down a socket.
1da177e4
LT
1710 */
1711
3e0fa65f
HC
1712SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1713 unsigned, flags)
1da177e4
LT
1714{
1715 return sys_sendto(fd, buff, len, flags, NULL, 0);
1716}
1717
1718/*
89bddce5 1719 * Receive a frame from the socket and optionally record the address of the
1da177e4
LT
1720 * sender. We verify the buffers are writable and if needed move the
1721 * sender address from kernel to user space.
1722 */
1723
3e0fa65f
HC
1724SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1725 unsigned, flags, struct sockaddr __user *, addr,
1726 int __user *, addr_len)
1da177e4
LT
1727{
1728 struct socket *sock;
1729 struct iovec iov;
1730 struct msghdr msg;
230b1839 1731 struct sockaddr_storage address;
89bddce5 1732 int err, err2;
6cb153ca
BL
1733 int fput_needed;
1734
253eacc0
LT
1735 if (size > INT_MAX)
1736 size = INT_MAX;
de0fa95c 1737 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4 1738 if (!sock)
de0fa95c 1739 goto out;
1da177e4 1740
89bddce5
SH
1741 msg.msg_control = NULL;
1742 msg.msg_controllen = 0;
1743 msg.msg_iovlen = 1;
1744 msg.msg_iov = &iov;
1745 iov.iov_len = size;
1746 iov.iov_base = ubuf;
230b1839
YH
1747 msg.msg_name = (struct sockaddr *)&address;
1748 msg.msg_namelen = sizeof(address);
1da177e4
LT
1749 if (sock->file->f_flags & O_NONBLOCK)
1750 flags |= MSG_DONTWAIT;
89bddce5 1751 err = sock_recvmsg(sock, &msg, size, flags);
1da177e4 1752
89bddce5 1753 if (err >= 0 && addr != NULL) {
230b1839
YH
1754 err2 = move_addr_to_user((struct sockaddr *)&address,
1755 msg.msg_namelen, addr, addr_len);
89bddce5
SH
1756 if (err2 < 0)
1757 err = err2;
1da177e4 1758 }
de0fa95c
PE
1759
1760 fput_light(sock->file, fput_needed);
4387ff75 1761out:
1da177e4
LT
1762 return err;
1763}
1764
1765/*
89bddce5 1766 * Receive a datagram from a socket.
1da177e4
LT
1767 */
1768
89bddce5
SH
1769asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1770 unsigned flags)
1da177e4
LT
1771{
1772 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1773}
1774
1775/*
1776 * Set a socket option. Because we don't know the option lengths we have
1777 * to pass the user mode parameter for the protocols to sort out.
1778 */
1779
20f37034
HC
1780SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1781 char __user *, optval, int, optlen)
1da177e4 1782{
6cb153ca 1783 int err, fput_needed;
1da177e4
LT
1784 struct socket *sock;
1785
1786 if (optlen < 0)
1787 return -EINVAL;
89bddce5
SH
1788
1789 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1790 if (sock != NULL) {
1791 err = security_socket_setsockopt(sock, level, optname);
6cb153ca
BL
1792 if (err)
1793 goto out_put;
1da177e4
LT
1794
1795 if (level == SOL_SOCKET)
89bddce5
SH
1796 err =
1797 sock_setsockopt(sock, level, optname, optval,
1798 optlen);
1da177e4 1799 else
89bddce5
SH
1800 err =
1801 sock->ops->setsockopt(sock, level, optname, optval,
1802 optlen);
6cb153ca
BL
1803out_put:
1804 fput_light(sock->file, fput_needed);
1da177e4
LT
1805 }
1806 return err;
1807}
1808
1809/*
1810 * Get a socket option. Because we don't know the option lengths we have
1811 * to pass a user mode parameter for the protocols to sort out.
1812 */
1813
20f37034
HC
1814SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1815 char __user *, optval, int __user *, optlen)
1da177e4 1816{
6cb153ca 1817 int err, fput_needed;
1da177e4
LT
1818 struct socket *sock;
1819
89bddce5
SH
1820 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1821 if (sock != NULL) {
6cb153ca
BL
1822 err = security_socket_getsockopt(sock, level, optname);
1823 if (err)
1824 goto out_put;
1da177e4
LT
1825
1826 if (level == SOL_SOCKET)
89bddce5
SH
1827 err =
1828 sock_getsockopt(sock, level, optname, optval,
1829 optlen);
1da177e4 1830 else
89bddce5
SH
1831 err =
1832 sock->ops->getsockopt(sock, level, optname, optval,
1833 optlen);
6cb153ca
BL
1834out_put:
1835 fput_light(sock->file, fput_needed);
1da177e4
LT
1836 }
1837 return err;
1838}
1839
1da177e4
LT
1840/*
1841 * Shutdown a socket.
1842 */
1843
754fe8d2 1844SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1da177e4 1845{
6cb153ca 1846 int err, fput_needed;
1da177e4
LT
1847 struct socket *sock;
1848
89bddce5
SH
1849 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1850 if (sock != NULL) {
1da177e4 1851 err = security_socket_shutdown(sock, how);
6cb153ca
BL
1852 if (!err)
1853 err = sock->ops->shutdown(sock, how);
1854 fput_light(sock->file, fput_needed);
1da177e4
LT
1855 }
1856 return err;
1857}
1858
89bddce5 1859/* A couple of helpful macros for getting the address of the 32/64 bit
1da177e4
LT
1860 * fields which are the same type (int / unsigned) on our platforms.
1861 */
1862#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1863#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1864#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1865
1da177e4
LT
1866/*
1867 * BSD sendmsg interface
1868 */
1869
3e0fa65f 1870SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1da177e4 1871{
89bddce5
SH
1872 struct compat_msghdr __user *msg_compat =
1873 (struct compat_msghdr __user *)msg;
1da177e4 1874 struct socket *sock;
230b1839 1875 struct sockaddr_storage address;
1da177e4 1876 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
b9d717a7 1877 unsigned char ctl[sizeof(struct cmsghdr) + 20]
89bddce5
SH
1878 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1879 /* 20 is size of ipv6_pktinfo */
1da177e4
LT
1880 unsigned char *ctl_buf = ctl;
1881 struct msghdr msg_sys;
1882 int err, ctl_len, iov_size, total_len;
6cb153ca 1883 int fput_needed;
89bddce5 1884
1da177e4
LT
1885 err = -EFAULT;
1886 if (MSG_CMSG_COMPAT & flags) {
1887 if (get_compat_msghdr(&msg_sys, msg_compat))
1888 return -EFAULT;
c6d409cf 1889 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1da177e4
LT
1890 return -EFAULT;
1891
6cb153ca 1892 sock = sockfd_lookup_light(fd, &err, &fput_needed);
89bddce5 1893 if (!sock)
1da177e4
LT
1894 goto out;
1895
1896 /* do not move before msg_sys is valid */
1897 err = -EMSGSIZE;
1898 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1899 goto out_put;
1900
89bddce5 1901 /* Check whether to allocate the iovec area */
1da177e4
LT
1902 err = -ENOMEM;
1903 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1904 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1905 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1906 if (!iov)
1907 goto out_put;
1908 }
1909
1910 /* This will also move the address data into kernel space */
1911 if (MSG_CMSG_COMPAT & flags) {
230b1839
YH
1912 err = verify_compat_iovec(&msg_sys, iov,
1913 (struct sockaddr *)&address,
1914 VERIFY_READ);
1da177e4 1915 } else
230b1839
YH
1916 err = verify_iovec(&msg_sys, iov,
1917 (struct sockaddr *)&address,
1918 VERIFY_READ);
89bddce5 1919 if (err < 0)
1da177e4
LT
1920 goto out_freeiov;
1921 total_len = err;
1922
1923 err = -ENOBUFS;
1924
1925 if (msg_sys.msg_controllen > INT_MAX)
1926 goto out_freeiov;
89bddce5 1927 ctl_len = msg_sys.msg_controllen;
1da177e4 1928 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
89bddce5
SH
1929 err =
1930 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1931 sizeof(ctl));
1da177e4
LT
1932 if (err)
1933 goto out_freeiov;
1934 ctl_buf = msg_sys.msg_control;
8920e8f9 1935 ctl_len = msg_sys.msg_controllen;
1da177e4 1936 } else if (ctl_len) {
89bddce5 1937 if (ctl_len > sizeof(ctl)) {
1da177e4 1938 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
89bddce5 1939 if (ctl_buf == NULL)
1da177e4
LT
1940 goto out_freeiov;
1941 }
1942 err = -EFAULT;
1943 /*
1944 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1945 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1946 * checking falls down on this.
1947 */
fb8621bb
NK
1948 if (copy_from_user(ctl_buf,
1949 (void __user __force *)msg_sys.msg_control,
89bddce5 1950 ctl_len))
1da177e4
LT
1951 goto out_freectl;
1952 msg_sys.msg_control = ctl_buf;
1953 }
1954 msg_sys.msg_flags = flags;
1955
1956 if (sock->file->f_flags & O_NONBLOCK)
1957 msg_sys.msg_flags |= MSG_DONTWAIT;
1958 err = sock_sendmsg(sock, &msg_sys, total_len);
1959
1960out_freectl:
89bddce5 1961 if (ctl_buf != ctl)
1da177e4
LT
1962 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1963out_freeiov:
1964 if (iov != iovstack)
1965 sock_kfree_s(sock->sk, iov, iov_size);
1966out_put:
6cb153ca 1967 fput_light(sock->file, fput_needed);
89bddce5 1968out:
1da177e4
LT
1969 return err;
1970}
1971
a2e27255
ACM
1972static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1973 struct msghdr *msg_sys, unsigned flags, int nosec)
1da177e4 1974{
89bddce5
SH
1975 struct compat_msghdr __user *msg_compat =
1976 (struct compat_msghdr __user *)msg;
1da177e4 1977 struct iovec iovstack[UIO_FASTIOV];
89bddce5 1978 struct iovec *iov = iovstack;
1da177e4
LT
1979 unsigned long cmsg_ptr;
1980 int err, iov_size, total_len, len;
1981
1982 /* kernel mode address */
230b1839 1983 struct sockaddr_storage addr;
1da177e4
LT
1984
1985 /* user mode address pointers */
1986 struct sockaddr __user *uaddr;
1987 int __user *uaddr_len;
89bddce5 1988
1da177e4 1989 if (MSG_CMSG_COMPAT & flags) {
a2e27255 1990 if (get_compat_msghdr(msg_sys, msg_compat))
1da177e4 1991 return -EFAULT;
c6d409cf 1992 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
89bddce5 1993 return -EFAULT;
1da177e4 1994
1da177e4 1995 err = -EMSGSIZE;
a2e27255
ACM
1996 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1997 goto out;
89bddce5
SH
1998
1999 /* Check whether to allocate the iovec area */
1da177e4 2000 err = -ENOMEM;
a2e27255
ACM
2001 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
2002 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1da177e4
LT
2003 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2004 if (!iov)
a2e27255 2005 goto out;
1da177e4
LT
2006 }
2007
2008 /*
89bddce5
SH
2009 * Save the user-mode address (verify_iovec will change the
2010 * kernel msghdr to use the kernel address space)
1da177e4 2011 */
89bddce5 2012
a2e27255 2013 uaddr = (__force void __user *)msg_sys->msg_name;
1da177e4
LT
2014 uaddr_len = COMPAT_NAMELEN(msg);
2015 if (MSG_CMSG_COMPAT & flags) {
a2e27255 2016 err = verify_compat_iovec(msg_sys, iov,
230b1839
YH
2017 (struct sockaddr *)&addr,
2018 VERIFY_WRITE);
1da177e4 2019 } else
a2e27255 2020 err = verify_iovec(msg_sys, iov,
230b1839
YH
2021 (struct sockaddr *)&addr,
2022 VERIFY_WRITE);
1da177e4
LT
2023 if (err < 0)
2024 goto out_freeiov;
89bddce5 2025 total_len = err;
1da177e4 2026
a2e27255
ACM
2027 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2028 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
89bddce5 2029
1da177e4
LT
2030 if (sock->file->f_flags & O_NONBLOCK)
2031 flags |= MSG_DONTWAIT;
a2e27255
ACM
2032 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2033 total_len, flags);
1da177e4
LT
2034 if (err < 0)
2035 goto out_freeiov;
2036 len = err;
2037
2038 if (uaddr != NULL) {
230b1839 2039 err = move_addr_to_user((struct sockaddr *)&addr,
a2e27255 2040 msg_sys->msg_namelen, uaddr,
89bddce5 2041 uaddr_len);
1da177e4
LT
2042 if (err < 0)
2043 goto out_freeiov;
2044 }
a2e27255 2045 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
37f7f421 2046 COMPAT_FLAGS(msg));
1da177e4
LT
2047 if (err)
2048 goto out_freeiov;
2049 if (MSG_CMSG_COMPAT & flags)
a2e27255 2050 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
1da177e4
LT
2051 &msg_compat->msg_controllen);
2052 else
a2e27255 2053 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
1da177e4
LT
2054 &msg->msg_controllen);
2055 if (err)
2056 goto out_freeiov;
2057 err = len;
2058
2059out_freeiov:
2060 if (iov != iovstack)
2061 sock_kfree_s(sock->sk, iov, iov_size);
a2e27255
ACM
2062out:
2063 return err;
2064}
2065
2066/*
2067 * BSD recvmsg interface
2068 */
2069
2070SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2071 unsigned int, flags)
2072{
2073 int fput_needed, err;
2074 struct msghdr msg_sys;
2075 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2076
2077 if (!sock)
2078 goto out;
2079
2080 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2081
6cb153ca 2082 fput_light(sock->file, fput_needed);
1da177e4
LT
2083out:
2084 return err;
2085}
2086
a2e27255
ACM
2087/*
2088 * Linux recvmmsg interface
2089 */
2090
2091int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2092 unsigned int flags, struct timespec *timeout)
2093{
2094 int fput_needed, err, datagrams;
2095 struct socket *sock;
2096 struct mmsghdr __user *entry;
d7256d0e 2097 struct compat_mmsghdr __user *compat_entry;
a2e27255
ACM
2098 struct msghdr msg_sys;
2099 struct timespec end_time;
2100
2101 if (timeout &&
2102 poll_select_set_timeout(&end_time, timeout->tv_sec,
2103 timeout->tv_nsec))
2104 return -EINVAL;
2105
2106 datagrams = 0;
2107
2108 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2109 if (!sock)
2110 return err;
2111
2112 err = sock_error(sock->sk);
2113 if (err)
2114 goto out_put;
2115
2116 entry = mmsg;
d7256d0e 2117 compat_entry = (struct compat_mmsghdr __user *)mmsg;
a2e27255
ACM
2118
2119 while (datagrams < vlen) {
2120 /*
2121 * No need to ask LSM for more than the first datagram.
2122 */
d7256d0e
JMG
2123 if (MSG_CMSG_COMPAT & flags) {
2124 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2125 &msg_sys, flags, datagrams);
2126 if (err < 0)
2127 break;
2128 err = __put_user(err, &compat_entry->msg_len);
2129 ++compat_entry;
2130 } else {
2131 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2132 &msg_sys, flags, datagrams);
2133 if (err < 0)
2134 break;
2135 err = put_user(err, &entry->msg_len);
2136 ++entry;
2137 }
2138
a2e27255
ACM
2139 if (err)
2140 break;
a2e27255
ACM
2141 ++datagrams;
2142
71c5c159
BB
2143 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2144 if (flags & MSG_WAITFORONE)
2145 flags |= MSG_DONTWAIT;
2146
a2e27255
ACM
2147 if (timeout) {
2148 ktime_get_ts(timeout);
2149 *timeout = timespec_sub(end_time, *timeout);
2150 if (timeout->tv_sec < 0) {
2151 timeout->tv_sec = timeout->tv_nsec = 0;
2152 break;
2153 }
2154
2155 /* Timeout, return less than vlen datagrams */
2156 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2157 break;
2158 }
2159
2160 /* Out of band data, return right away */
2161 if (msg_sys.msg_flags & MSG_OOB)
2162 break;
2163 }
2164
2165out_put:
2166 fput_light(sock->file, fput_needed);
1da177e4 2167
a2e27255
ACM
2168 if (err == 0)
2169 return datagrams;
2170
2171 if (datagrams != 0) {
2172 /*
2173 * We may return less entries than requested (vlen) if the
2174 * sock is non block and there aren't enough datagrams...
2175 */
2176 if (err != -EAGAIN) {
2177 /*
2178 * ... or if recvmsg returns an error after we
2179 * received some datagrams, where we record the
2180 * error to return on the next call or if the
2181 * app asks about it using getsockopt(SO_ERROR).
2182 */
2183 sock->sk->sk_err = -err;
2184 }
2185
2186 return datagrams;
2187 }
2188
2189 return err;
2190}
2191
2192SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2193 unsigned int, vlen, unsigned int, flags,
2194 struct timespec __user *, timeout)
2195{
2196 int datagrams;
2197 struct timespec timeout_sys;
2198
2199 if (!timeout)
2200 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2201
2202 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2203 return -EFAULT;
2204
2205 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2206
2207 if (datagrams > 0 &&
2208 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2209 datagrams = -EFAULT;
2210
2211 return datagrams;
2212}
2213
2214#ifdef __ARCH_WANT_SYS_SOCKETCALL
1da177e4
LT
2215/* Argument list sizes for sys_socketcall */
2216#define AL(x) ((x) * sizeof(unsigned long))
a2e27255 2217static const unsigned char nargs[20] = {
c6d409cf
ED
2218 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2219 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2220 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2221 AL(4), AL(5)
89bddce5
SH
2222};
2223
1da177e4
LT
2224#undef AL
2225
2226/*
89bddce5 2227 * System call vectors.
1da177e4
LT
2228 *
2229 * Argument checking cleaned up. Saved 20% in size.
2230 * This function doesn't need to set the kernel lock because
89bddce5 2231 * it is set by the callees.
1da177e4
LT
2232 */
2233
3e0fa65f 2234SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
1da177e4
LT
2235{
2236 unsigned long a[6];
89bddce5 2237 unsigned long a0, a1;
1da177e4 2238 int err;
47379052 2239 unsigned int len;
1da177e4 2240
a2e27255 2241 if (call < 1 || call > SYS_RECVMMSG)
1da177e4
LT
2242 return -EINVAL;
2243
47379052
AV
2244 len = nargs[call];
2245 if (len > sizeof(a))
2246 return -EINVAL;
2247
1da177e4 2248 /* copy_from_user should be SMP safe. */
47379052 2249 if (copy_from_user(a, args, len))
1da177e4 2250 return -EFAULT;
3ec3b2fb 2251
f3298dc4 2252 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3ec3b2fb 2253
89bddce5
SH
2254 a0 = a[0];
2255 a1 = a[1];
2256
2257 switch (call) {
2258 case SYS_SOCKET:
2259 err = sys_socket(a0, a1, a[2]);
2260 break;
2261 case SYS_BIND:
2262 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2263 break;
2264 case SYS_CONNECT:
2265 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2266 break;
2267 case SYS_LISTEN:
2268 err = sys_listen(a0, a1);
2269 break;
2270 case SYS_ACCEPT:
de11defe
UD
2271 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2272 (int __user *)a[2], 0);
89bddce5
SH
2273 break;
2274 case SYS_GETSOCKNAME:
2275 err =
2276 sys_getsockname(a0, (struct sockaddr __user *)a1,
2277 (int __user *)a[2]);
2278 break;
2279 case SYS_GETPEERNAME:
2280 err =
2281 sys_getpeername(a0, (struct sockaddr __user *)a1,
2282 (int __user *)a[2]);
2283 break;
2284 case SYS_SOCKETPAIR:
2285 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2286 break;
2287 case SYS_SEND:
2288 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2289 break;
2290 case SYS_SENDTO:
2291 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2292 (struct sockaddr __user *)a[4], a[5]);
2293 break;
2294 case SYS_RECV:
2295 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2296 break;
2297 case SYS_RECVFROM:
2298 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2299 (struct sockaddr __user *)a[4],
2300 (int __user *)a[5]);
2301 break;
2302 case SYS_SHUTDOWN:
2303 err = sys_shutdown(a0, a1);
2304 break;
2305 case SYS_SETSOCKOPT:
2306 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2307 break;
2308 case SYS_GETSOCKOPT:
2309 err =
2310 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2311 (int __user *)a[4]);
2312 break;
2313 case SYS_SENDMSG:
2314 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2315 break;
2316 case SYS_RECVMSG:
2317 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2318 break;
a2e27255
ACM
2319 case SYS_RECVMMSG:
2320 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2321 (struct timespec __user *)a[4]);
2322 break;
de11defe
UD
2323 case SYS_ACCEPT4:
2324 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2325 (int __user *)a[2], a[3]);
aaca0bdc 2326 break;
89bddce5
SH
2327 default:
2328 err = -EINVAL;
2329 break;
1da177e4
LT
2330 }
2331 return err;
2332}
2333
89bddce5 2334#endif /* __ARCH_WANT_SYS_SOCKETCALL */
1da177e4 2335
55737fda
SH
2336/**
2337 * sock_register - add a socket protocol handler
2338 * @ops: description of protocol
2339 *
1da177e4
LT
2340 * This function is called by a protocol handler that wants to
2341 * advertise its address family, and have it linked into the
55737fda
SH
2342 * socket interface. The value ops->family coresponds to the
2343 * socket system call protocol family.
1da177e4 2344 */
f0fd27d4 2345int sock_register(const struct net_proto_family *ops)
1da177e4
LT
2346{
2347 int err;
2348
2349 if (ops->family >= NPROTO) {
89bddce5
SH
2350 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2351 NPROTO);
1da177e4
LT
2352 return -ENOBUFS;
2353 }
55737fda
SH
2354
2355 spin_lock(&net_family_lock);
190683a9
ED
2356 if (rcu_dereference_protected(net_families[ops->family],
2357 lockdep_is_held(&net_family_lock)))
55737fda
SH
2358 err = -EEXIST;
2359 else {
190683a9 2360 rcu_assign_pointer(net_families[ops->family], ops);
1da177e4
LT
2361 err = 0;
2362 }
55737fda
SH
2363 spin_unlock(&net_family_lock);
2364
89bddce5 2365 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
1da177e4
LT
2366 return err;
2367}
c6d409cf 2368EXPORT_SYMBOL(sock_register);
1da177e4 2369
55737fda
SH
2370/**
2371 * sock_unregister - remove a protocol handler
2372 * @family: protocol family to remove
2373 *
1da177e4
LT
2374 * This function is called by a protocol handler that wants to
2375 * remove its address family, and have it unlinked from the
55737fda
SH
2376 * new socket creation.
2377 *
2378 * If protocol handler is a module, then it can use module reference
2379 * counts to protect against new references. If protocol handler is not
2380 * a module then it needs to provide its own protection in
2381 * the ops->create routine.
1da177e4 2382 */
f0fd27d4 2383void sock_unregister(int family)
1da177e4 2384{
f0fd27d4 2385 BUG_ON(family < 0 || family >= NPROTO);
1da177e4 2386
55737fda 2387 spin_lock(&net_family_lock);
190683a9 2388 rcu_assign_pointer(net_families[family], NULL);
55737fda
SH
2389 spin_unlock(&net_family_lock);
2390
2391 synchronize_rcu();
2392
89bddce5 2393 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
1da177e4 2394}
c6d409cf 2395EXPORT_SYMBOL(sock_unregister);
1da177e4 2396
77d76ea3 2397static int __init sock_init(void)
1da177e4 2398{
b3e19d92
NP
2399 int err;
2400
1da177e4 2401 /*
89bddce5 2402 * Initialize sock SLAB cache.
1da177e4 2403 */
89bddce5 2404
1da177e4
LT
2405 sk_init();
2406
1da177e4 2407 /*
89bddce5 2408 * Initialize skbuff SLAB cache
1da177e4
LT
2409 */
2410 skb_init();
1da177e4
LT
2411
2412 /*
89bddce5 2413 * Initialize the protocols module.
1da177e4
LT
2414 */
2415
2416 init_inodecache();
b3e19d92
NP
2417
2418 err = register_filesystem(&sock_fs_type);
2419 if (err)
2420 goto out_fs;
1da177e4 2421 sock_mnt = kern_mount(&sock_fs_type);
b3e19d92
NP
2422 if (IS_ERR(sock_mnt)) {
2423 err = PTR_ERR(sock_mnt);
2424 goto out_mount;
2425 }
77d76ea3
AK
2426
2427 /* The real protocol initialization is performed in later initcalls.
1da177e4
LT
2428 */
2429
2430#ifdef CONFIG_NETFILTER
2431 netfilter_init();
2432#endif
cbeb321a 2433
c1f19b51
RC
2434#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2435 skb_timestamping_init();
2436#endif
2437
b3e19d92
NP
2438out:
2439 return err;
2440
2441out_mount:
2442 unregister_filesystem(&sock_fs_type);
2443out_fs:
2444 goto out;
1da177e4
LT
2445}
2446
77d76ea3
AK
2447core_initcall(sock_init); /* early initcall */
2448
1da177e4
LT
2449#ifdef CONFIG_PROC_FS
2450void socket_seq_show(struct seq_file *seq)
2451{
2452 int cpu;
2453 int counter = 0;
2454
6f912042 2455 for_each_possible_cpu(cpu)
89bddce5 2456 counter += per_cpu(sockets_in_use, cpu);
1da177e4
LT
2457
2458 /* It can be negative, by the way. 8) */
2459 if (counter < 0)
2460 counter = 0;
2461
2462 seq_printf(seq, "sockets: used %d\n", counter);
2463}
89bddce5 2464#endif /* CONFIG_PROC_FS */
1da177e4 2465
89bbfc95 2466#ifdef CONFIG_COMPAT
6b96018b
AB
2467static int do_siocgstamp(struct net *net, struct socket *sock,
2468 unsigned int cmd, struct compat_timeval __user *up)
7a229387 2469{
7a229387
AB
2470 mm_segment_t old_fs = get_fs();
2471 struct timeval ktv;
2472 int err;
2473
2474 set_fs(KERNEL_DS);
6b96018b 2475 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
7a229387
AB
2476 set_fs(old_fs);
2477 if (!err) {
2478 err = put_user(ktv.tv_sec, &up->tv_sec);
2479 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2480 }
2481 return err;
2482}
2483
6b96018b
AB
2484static int do_siocgstampns(struct net *net, struct socket *sock,
2485 unsigned int cmd, struct compat_timespec __user *up)
7a229387 2486{
7a229387
AB
2487 mm_segment_t old_fs = get_fs();
2488 struct timespec kts;
2489 int err;
2490
2491 set_fs(KERNEL_DS);
6b96018b 2492 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
7a229387
AB
2493 set_fs(old_fs);
2494 if (!err) {
2495 err = put_user(kts.tv_sec, &up->tv_sec);
2496 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2497 }
2498 return err;
2499}
2500
6b96018b 2501static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
7a229387
AB
2502{
2503 struct ifreq __user *uifr;
2504 int err;
2505
2506 uifr = compat_alloc_user_space(sizeof(struct ifreq));
6b96018b 2507 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
7a229387
AB
2508 return -EFAULT;
2509
6b96018b 2510 err = dev_ioctl(net, SIOCGIFNAME, uifr);
7a229387
AB
2511 if (err)
2512 return err;
2513
6b96018b 2514 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
7a229387
AB
2515 return -EFAULT;
2516
2517 return 0;
2518}
2519
6b96018b 2520static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
7a229387 2521{
6b96018b 2522 struct compat_ifconf ifc32;
7a229387
AB
2523 struct ifconf ifc;
2524 struct ifconf __user *uifc;
6b96018b 2525 struct compat_ifreq __user *ifr32;
7a229387
AB
2526 struct ifreq __user *ifr;
2527 unsigned int i, j;
2528 int err;
2529
6b96018b 2530 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
7a229387
AB
2531 return -EFAULT;
2532
2533 if (ifc32.ifcbuf == 0) {
2534 ifc32.ifc_len = 0;
2535 ifc.ifc_len = 0;
2536 ifc.ifc_req = NULL;
2537 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2538 } else {
c6d409cf
ED
2539 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2540 sizeof(struct ifreq);
7a229387
AB
2541 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2542 ifc.ifc_len = len;
2543 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2544 ifr32 = compat_ptr(ifc32.ifcbuf);
c6d409cf 2545 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
6b96018b 2546 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
7a229387
AB
2547 return -EFAULT;
2548 ifr++;
2549 ifr32++;
2550 }
2551 }
2552 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2553 return -EFAULT;
2554
6b96018b 2555 err = dev_ioctl(net, SIOCGIFCONF, uifc);
7a229387
AB
2556 if (err)
2557 return err;
2558
2559 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2560 return -EFAULT;
2561
2562 ifr = ifc.ifc_req;
2563 ifr32 = compat_ptr(ifc32.ifcbuf);
2564 for (i = 0, j = 0;
c6d409cf
ED
2565 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2566 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2567 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
7a229387
AB
2568 return -EFAULT;
2569 ifr32++;
2570 ifr++;
2571 }
2572
2573 if (ifc32.ifcbuf == 0) {
2574 /* Translate from 64-bit structure multiple to
2575 * a 32-bit one.
2576 */
2577 i = ifc.ifc_len;
6b96018b 2578 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
7a229387
AB
2579 ifc32.ifc_len = i;
2580 } else {
2581 ifc32.ifc_len = i;
2582 }
6b96018b 2583 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
7a229387
AB
2584 return -EFAULT;
2585
2586 return 0;
2587}
2588
6b96018b 2589static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
7a229387
AB
2590{
2591 struct ifreq __user *ifr;
7a229387
AB
2592 u32 data;
2593 void __user *datap;
2594
2595 ifr = compat_alloc_user_space(sizeof(*ifr));
7a229387
AB
2596
2597 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2598 return -EFAULT;
2599
2600 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2601 return -EFAULT;
2602
2603 datap = compat_ptr(data);
2604 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2605 return -EFAULT;
2606
6b96018b 2607 return dev_ioctl(net, SIOCETHTOOL, ifr);
7a229387
AB
2608}
2609
7a50a240
AB
2610static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2611{
2612 void __user *uptr;
2613 compat_uptr_t uptr32;
2614 struct ifreq __user *uifr;
2615
c6d409cf 2616 uifr = compat_alloc_user_space(sizeof(*uifr));
7a50a240
AB
2617 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2618 return -EFAULT;
2619
2620 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2621 return -EFAULT;
2622
2623 uptr = compat_ptr(uptr32);
2624
2625 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2626 return -EFAULT;
2627
2628 return dev_ioctl(net, SIOCWANDEV, uifr);
2629}
2630
6b96018b
AB
2631static int bond_ioctl(struct net *net, unsigned int cmd,
2632 struct compat_ifreq __user *ifr32)
7a229387
AB
2633{
2634 struct ifreq kifr;
2635 struct ifreq __user *uifr;
7a229387
AB
2636 mm_segment_t old_fs;
2637 int err;
2638 u32 data;
2639 void __user *datap;
2640
2641 switch (cmd) {
2642 case SIOCBONDENSLAVE:
2643 case SIOCBONDRELEASE:
2644 case SIOCBONDSETHWADDR:
2645 case SIOCBONDCHANGEACTIVE:
6b96018b 2646 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
7a229387
AB
2647 return -EFAULT;
2648
2649 old_fs = get_fs();
c6d409cf 2650 set_fs(KERNEL_DS);
6b96018b 2651 err = dev_ioctl(net, cmd, &kifr);
c6d409cf 2652 set_fs(old_fs);
7a229387
AB
2653
2654 return err;
2655 case SIOCBONDSLAVEINFOQUERY:
2656 case SIOCBONDINFOQUERY:
2657 uifr = compat_alloc_user_space(sizeof(*uifr));
2658 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2659 return -EFAULT;
2660
2661 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2662 return -EFAULT;
2663
2664 datap = compat_ptr(data);
2665 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2666 return -EFAULT;
2667
6b96018b 2668 return dev_ioctl(net, cmd, uifr);
7a229387
AB
2669 default:
2670 return -EINVAL;
ccbd6a5a 2671 }
7a229387
AB
2672}
2673
6b96018b
AB
2674static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2675 struct compat_ifreq __user *u_ifreq32)
7a229387
AB
2676{
2677 struct ifreq __user *u_ifreq64;
7a229387
AB
2678 char tmp_buf[IFNAMSIZ];
2679 void __user *data64;
2680 u32 data32;
2681
2682 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2683 IFNAMSIZ))
2684 return -EFAULT;
2685 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2686 return -EFAULT;
2687 data64 = compat_ptr(data32);
2688
2689 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2690
2691 /* Don't check these user accesses, just let that get trapped
2692 * in the ioctl handler instead.
2693 */
2694 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2695 IFNAMSIZ))
2696 return -EFAULT;
2697 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2698 return -EFAULT;
2699
6b96018b 2700 return dev_ioctl(net, cmd, u_ifreq64);
7a229387
AB
2701}
2702
6b96018b
AB
2703static int dev_ifsioc(struct net *net, struct socket *sock,
2704 unsigned int cmd, struct compat_ifreq __user *uifr32)
7a229387 2705{
a2116ed2 2706 struct ifreq __user *uifr;
7a229387
AB
2707 int err;
2708
a2116ed2
AB
2709 uifr = compat_alloc_user_space(sizeof(*uifr));
2710 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2711 return -EFAULT;
2712
2713 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2714
7a229387
AB
2715 if (!err) {
2716 switch (cmd) {
2717 case SIOCGIFFLAGS:
2718 case SIOCGIFMETRIC:
2719 case SIOCGIFMTU:
2720 case SIOCGIFMEM:
2721 case SIOCGIFHWADDR:
2722 case SIOCGIFINDEX:
2723 case SIOCGIFADDR:
2724 case SIOCGIFBRDADDR:
2725 case SIOCGIFDSTADDR:
2726 case SIOCGIFNETMASK:
fab2532b 2727 case SIOCGIFPFLAGS:
7a229387 2728 case SIOCGIFTXQLEN:
fab2532b
AB
2729 case SIOCGMIIPHY:
2730 case SIOCGMIIREG:
a2116ed2 2731 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
7a229387
AB
2732 err = -EFAULT;
2733 break;
2734 }
2735 }
2736 return err;
2737}
2738
a2116ed2
AB
2739static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2740 struct compat_ifreq __user *uifr32)
2741{
2742 struct ifreq ifr;
2743 struct compat_ifmap __user *uifmap32;
2744 mm_segment_t old_fs;
2745 int err;
2746
2747 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2748 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2749 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2750 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2751 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2752 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2753 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2754 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2755 if (err)
2756 return -EFAULT;
2757
2758 old_fs = get_fs();
c6d409cf 2759 set_fs(KERNEL_DS);
a2116ed2 2760 err = dev_ioctl(net, cmd, (void __user *)&ifr);
c6d409cf 2761 set_fs(old_fs);
a2116ed2
AB
2762
2763 if (cmd == SIOCGIFMAP && !err) {
2764 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2765 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2766 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2767 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2768 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2769 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2770 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2771 if (err)
2772 err = -EFAULT;
2773 }
2774 return err;
2775}
2776
2777static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2778{
2779 void __user *uptr;
2780 compat_uptr_t uptr32;
2781 struct ifreq __user *uifr;
2782
c6d409cf 2783 uifr = compat_alloc_user_space(sizeof(*uifr));
a2116ed2
AB
2784 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2785 return -EFAULT;
2786
2787 if (get_user(uptr32, &uifr32->ifr_data))
2788 return -EFAULT;
2789
2790 uptr = compat_ptr(uptr32);
2791
2792 if (put_user(uptr, &uifr->ifr_data))
2793 return -EFAULT;
2794
2795 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2796}
2797
7a229387 2798struct rtentry32 {
c6d409cf 2799 u32 rt_pad1;
7a229387
AB
2800 struct sockaddr rt_dst; /* target address */
2801 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2802 struct sockaddr rt_genmask; /* target network mask (IP) */
c6d409cf
ED
2803 unsigned short rt_flags;
2804 short rt_pad2;
2805 u32 rt_pad3;
2806 unsigned char rt_tos;
2807 unsigned char rt_class;
2808 short rt_pad4;
2809 short rt_metric; /* +1 for binary compatibility! */
7a229387 2810 /* char * */ u32 rt_dev; /* forcing the device at add */
c6d409cf
ED
2811 u32 rt_mtu; /* per route MTU/Window */
2812 u32 rt_window; /* Window clamping */
7a229387
AB
2813 unsigned short rt_irtt; /* Initial RTT */
2814};
2815
2816struct in6_rtmsg32 {
2817 struct in6_addr rtmsg_dst;
2818 struct in6_addr rtmsg_src;
2819 struct in6_addr rtmsg_gateway;
2820 u32 rtmsg_type;
2821 u16 rtmsg_dst_len;
2822 u16 rtmsg_src_len;
2823 u32 rtmsg_metric;
2824 u32 rtmsg_info;
2825 u32 rtmsg_flags;
2826 s32 rtmsg_ifindex;
2827};
2828
6b96018b
AB
2829static int routing_ioctl(struct net *net, struct socket *sock,
2830 unsigned int cmd, void __user *argp)
7a229387
AB
2831{
2832 int ret;
2833 void *r = NULL;
2834 struct in6_rtmsg r6;
2835 struct rtentry r4;
2836 char devname[16];
2837 u32 rtdev;
2838 mm_segment_t old_fs = get_fs();
2839
6b96018b
AB
2840 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2841 struct in6_rtmsg32 __user *ur6 = argp;
c6d409cf 2842 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
7a229387 2843 3 * sizeof(struct in6_addr));
c6d409cf
ED
2844 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2845 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2846 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2847 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2848 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2849 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2850 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
7a229387
AB
2851
2852 r = (void *) &r6;
2853 } else { /* ipv4 */
6b96018b 2854 struct rtentry32 __user *ur4 = argp;
c6d409cf 2855 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
7a229387 2856 3 * sizeof(struct sockaddr));
c6d409cf
ED
2857 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
2858 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
2859 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
2860 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
2861 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
2862 ret |= __get_user(rtdev, &(ur4->rt_dev));
7a229387 2863 if (rtdev) {
c6d409cf 2864 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
7a229387
AB
2865 r4.rt_dev = devname; devname[15] = 0;
2866 } else
2867 r4.rt_dev = NULL;
2868
2869 r = (void *) &r4;
2870 }
2871
2872 if (ret) {
2873 ret = -EFAULT;
2874 goto out;
2875 }
2876
c6d409cf 2877 set_fs(KERNEL_DS);
6b96018b 2878 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
c6d409cf 2879 set_fs(old_fs);
7a229387
AB
2880
2881out:
7a229387
AB
2882 return ret;
2883}
2884
2885/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2886 * for some operations; this forces use of the newer bridge-utils that
2887 * use compatiable ioctls
2888 */
6b96018b 2889static int old_bridge_ioctl(compat_ulong_t __user *argp)
7a229387 2890{
6b96018b 2891 compat_ulong_t tmp;
7a229387 2892
6b96018b 2893 if (get_user(tmp, argp))
7a229387
AB
2894 return -EFAULT;
2895 if (tmp == BRCTL_GET_VERSION)
2896 return BRCTL_VERSION + 1;
2897 return -EINVAL;
2898}
2899
6b96018b
AB
2900static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2901 unsigned int cmd, unsigned long arg)
2902{
2903 void __user *argp = compat_ptr(arg);
2904 struct sock *sk = sock->sk;
2905 struct net *net = sock_net(sk);
7a229387 2906
6b96018b
AB
2907 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2908 return siocdevprivate_ioctl(net, cmd, argp);
2909
2910 switch (cmd) {
2911 case SIOCSIFBR:
2912 case SIOCGIFBR:
2913 return old_bridge_ioctl(argp);
2914 case SIOCGIFNAME:
2915 return dev_ifname32(net, argp);
2916 case SIOCGIFCONF:
2917 return dev_ifconf(net, argp);
2918 case SIOCETHTOOL:
2919 return ethtool_ioctl(net, argp);
7a50a240
AB
2920 case SIOCWANDEV:
2921 return compat_siocwandev(net, argp);
a2116ed2
AB
2922 case SIOCGIFMAP:
2923 case SIOCSIFMAP:
2924 return compat_sioc_ifmap(net, cmd, argp);
6b96018b
AB
2925 case SIOCBONDENSLAVE:
2926 case SIOCBONDRELEASE:
2927 case SIOCBONDSETHWADDR:
2928 case SIOCBONDSLAVEINFOQUERY:
2929 case SIOCBONDINFOQUERY:
2930 case SIOCBONDCHANGEACTIVE:
2931 return bond_ioctl(net, cmd, argp);
2932 case SIOCADDRT:
2933 case SIOCDELRT:
2934 return routing_ioctl(net, sock, cmd, argp);
2935 case SIOCGSTAMP:
2936 return do_siocgstamp(net, sock, cmd, argp);
2937 case SIOCGSTAMPNS:
2938 return do_siocgstampns(net, sock, cmd, argp);
a2116ed2
AB
2939 case SIOCSHWTSTAMP:
2940 return compat_siocshwtstamp(net, argp);
6b96018b
AB
2941
2942 case FIOSETOWN:
2943 case SIOCSPGRP:
2944 case FIOGETOWN:
2945 case SIOCGPGRP:
2946 case SIOCBRADDBR:
2947 case SIOCBRDELBR:
2948 case SIOCGIFVLAN:
2949 case SIOCSIFVLAN:
2950 case SIOCADDDLCI:
2951 case SIOCDELDLCI:
2952 return sock_ioctl(file, cmd, arg);
2953
2954 case SIOCGIFFLAGS:
2955 case SIOCSIFFLAGS:
2956 case SIOCGIFMETRIC:
2957 case SIOCSIFMETRIC:
2958 case SIOCGIFMTU:
2959 case SIOCSIFMTU:
2960 case SIOCGIFMEM:
2961 case SIOCSIFMEM:
2962 case SIOCGIFHWADDR:
2963 case SIOCSIFHWADDR:
2964 case SIOCADDMULTI:
2965 case SIOCDELMULTI:
2966 case SIOCGIFINDEX:
6b96018b
AB
2967 case SIOCGIFADDR:
2968 case SIOCSIFADDR:
2969 case SIOCSIFHWBROADCAST:
6b96018b 2970 case SIOCDIFADDR:
6b96018b
AB
2971 case SIOCGIFBRDADDR:
2972 case SIOCSIFBRDADDR:
2973 case SIOCGIFDSTADDR:
2974 case SIOCSIFDSTADDR:
2975 case SIOCGIFNETMASK:
2976 case SIOCSIFNETMASK:
2977 case SIOCSIFPFLAGS:
2978 case SIOCGIFPFLAGS:
2979 case SIOCGIFTXQLEN:
2980 case SIOCSIFTXQLEN:
2981 case SIOCBRADDIF:
2982 case SIOCBRDELIF:
9177efd3
AB
2983 case SIOCSIFNAME:
2984 case SIOCGMIIPHY:
2985 case SIOCGMIIREG:
2986 case SIOCSMIIREG:
6b96018b 2987 return dev_ifsioc(net, sock, cmd, argp);
9177efd3 2988
6b96018b
AB
2989 case SIOCSARP:
2990 case SIOCGARP:
2991 case SIOCDARP:
6b96018b 2992 case SIOCATMARK:
9177efd3
AB
2993 return sock_do_ioctl(net, sock, cmd, arg);
2994 }
2995
2996 /* Prevent warning from compat_sys_ioctl, these always
2997 * result in -EINVAL in the native case anyway. */
2998 switch (cmd) {
2999 case SIOCRTMSG:
3000 case SIOCGIFCOUNT:
6b96018b
AB
3001 case SIOCSRARP:
3002 case SIOCGRARP:
3003 case SIOCDRARP:
9177efd3
AB
3004 case SIOCSIFLINK:
3005 case SIOCGIFSLAVE:
3006 case SIOCSIFSLAVE:
3007 return -EINVAL;
6b96018b
AB
3008 }
3009
3010 return -ENOIOCTLCMD;
3011}
7a229387 3012
89bbfc95 3013static long compat_sock_ioctl(struct file *file, unsigned cmd,
89bddce5 3014 unsigned long arg)
89bbfc95
SP
3015{
3016 struct socket *sock = file->private_data;
3017 int ret = -ENOIOCTLCMD;
87de87d5
DM
3018 struct sock *sk;
3019 struct net *net;
3020
3021 sk = sock->sk;
3022 net = sock_net(sk);
89bbfc95
SP
3023
3024 if (sock->ops->compat_ioctl)
3025 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3026
87de87d5
DM
3027 if (ret == -ENOIOCTLCMD &&
3028 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3029 ret = compat_wext_handle_ioctl(net, cmd, arg);
3030
6b96018b
AB
3031 if (ret == -ENOIOCTLCMD)
3032 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3033
89bbfc95
SP
3034 return ret;
3035}
3036#endif
3037
ac5a488e
SS
3038int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3039{
3040 return sock->ops->bind(sock, addr, addrlen);
3041}
c6d409cf 3042EXPORT_SYMBOL(kernel_bind);
ac5a488e
SS
3043
3044int kernel_listen(struct socket *sock, int backlog)
3045{
3046 return sock->ops->listen(sock, backlog);
3047}
c6d409cf 3048EXPORT_SYMBOL(kernel_listen);
ac5a488e
SS
3049
3050int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3051{
3052 struct sock *sk = sock->sk;
3053 int err;
3054
3055 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3056 newsock);
3057 if (err < 0)
3058 goto done;
3059
3060 err = sock->ops->accept(sock, *newsock, flags);
3061 if (err < 0) {
3062 sock_release(*newsock);
fa8705b0 3063 *newsock = NULL;
ac5a488e
SS
3064 goto done;
3065 }
3066
3067 (*newsock)->ops = sock->ops;
1b08534e 3068 __module_get((*newsock)->ops->owner);
ac5a488e
SS
3069
3070done:
3071 return err;
3072}
c6d409cf 3073EXPORT_SYMBOL(kernel_accept);
ac5a488e
SS
3074
3075int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
4768fbcb 3076 int flags)
ac5a488e
SS
3077{
3078 return sock->ops->connect(sock, addr, addrlen, flags);
3079}
c6d409cf 3080EXPORT_SYMBOL(kernel_connect);
ac5a488e
SS
3081
3082int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3083 int *addrlen)
3084{
3085 return sock->ops->getname(sock, addr, addrlen, 0);
3086}
c6d409cf 3087EXPORT_SYMBOL(kernel_getsockname);
ac5a488e
SS
3088
3089int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3090 int *addrlen)
3091{
3092 return sock->ops->getname(sock, addr, addrlen, 1);
3093}
c6d409cf 3094EXPORT_SYMBOL(kernel_getpeername);
ac5a488e
SS
3095
3096int kernel_getsockopt(struct socket *sock, int level, int optname,
3097 char *optval, int *optlen)
3098{
3099 mm_segment_t oldfs = get_fs();
fb8621bb
NK
3100 char __user *uoptval;
3101 int __user *uoptlen;
ac5a488e
SS
3102 int err;
3103
fb8621bb
NK
3104 uoptval = (char __user __force *) optval;
3105 uoptlen = (int __user __force *) optlen;
3106
ac5a488e
SS
3107 set_fs(KERNEL_DS);
3108 if (level == SOL_SOCKET)
fb8621bb 3109 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
ac5a488e 3110 else
fb8621bb
NK
3111 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3112 uoptlen);
ac5a488e
SS
3113 set_fs(oldfs);
3114 return err;
3115}
c6d409cf 3116EXPORT_SYMBOL(kernel_getsockopt);
ac5a488e
SS
3117
3118int kernel_setsockopt(struct socket *sock, int level, int optname,
b7058842 3119 char *optval, unsigned int optlen)
ac5a488e
SS
3120{
3121 mm_segment_t oldfs = get_fs();
fb8621bb 3122 char __user *uoptval;
ac5a488e
SS
3123 int err;
3124
fb8621bb
NK
3125 uoptval = (char __user __force *) optval;
3126
ac5a488e
SS
3127 set_fs(KERNEL_DS);
3128 if (level == SOL_SOCKET)
fb8621bb 3129 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
ac5a488e 3130 else
fb8621bb 3131 err = sock->ops->setsockopt(sock, level, optname, uoptval,
ac5a488e
SS
3132 optlen);
3133 set_fs(oldfs);
3134 return err;
3135}
c6d409cf 3136EXPORT_SYMBOL(kernel_setsockopt);
ac5a488e
SS
3137
3138int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3139 size_t size, int flags)
3140{
f8451725
HX
3141 sock_update_classid(sock->sk);
3142
ac5a488e
SS
3143 if (sock->ops->sendpage)
3144 return sock->ops->sendpage(sock, page, offset, size, flags);
3145
3146 return sock_no_sendpage(sock, page, offset, size, flags);
3147}
c6d409cf 3148EXPORT_SYMBOL(kernel_sendpage);
ac5a488e
SS
3149
3150int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3151{
3152 mm_segment_t oldfs = get_fs();
3153 int err;
3154
3155 set_fs(KERNEL_DS);
3156 err = sock->ops->ioctl(sock, cmd, arg);
3157 set_fs(oldfs);
3158
3159 return err;
3160}
c6d409cf 3161EXPORT_SYMBOL(kernel_sock_ioctl);
ac5a488e 3162
91cf45f0
TM
3163int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3164{
3165 return sock->ops->shutdown(sock, how);
3166}
91cf45f0 3167EXPORT_SYMBOL(kernel_sock_shutdown);